Items prohibited by law from being copied or reproduced Do not copy or reproduce paper money, coins, securities, government bonds, or local government bonds, even if such copies or reproductions are stamped Sample. The copying or reproduction of paper money, coins, or securities which are circulated in a foreign country is prohibited. Unless the prior permission of the government has been obtained, the copying or reproduction of unused postage stamps or post cards issued by the government is prohibited. The copying or reproduction of stamps issued by the government and of certified documents stipulated by law is prohibited. Cautions on certain copies and reproductions The government has issued cautions on copies or reproductions of securities issued by private companies (shares, bills, checks, gift certificates, etc.), commuter passes, or coupon tickets, except when a minimum of necessary copies are to be provided for business use by a company. Also, do not copy or reproduce passports issued by the government, licenses issued by public agencies and private groups, ID cards, and tickets, such as passes and meal coupons.
Comply with copyright notices The copying or reproduction of copyrighted creative works such as books, music, paintings, woodcut prints, maps, drawings, movies, and photographs is governed by national and international copyright laws. Do not use this product for the purpose of making illegal copies or to infringe copyright laws.
Disposing of Data Storage Devices Please note that deleting images or formatting data storage devices such as memory cards or built-in camera memory does not completely erase the original image data. Deleted files can sometimes be recovered from discarded storage devices using commercially available software, potentially resulting in the malicious use of personal image data. Ensuring the privacy of such data is the users responsibility. Before discarding a data storage device or transferring ownership to another person, erase all data using commercial deletion software, or format the device and then completely refill it with images containing no private information (for example, pictures of empty sky). Be sure to also replace any pictures selected for the welcome screen ( 57). Care should be taken to avoid injury or damage to property when physically destroying data storage devices.

Table of Contents

For Your Safety.... ii

WARNINGS..... ii

Parts of the Camera

Shutter-release button ( 16) Power switch ( 9) Power-on lamp (
9) Self-timer lamp ( 20) Built-in flash ( 18)

Eyelet for camera strap

Lens (

14, 78)

Lens cover closed
Power cord channel cover for optional AC adapter ( 76) Battery-chamber cover (
Lens cover Battery chamber ( 6)
Attaching the Camera Strap Attach the strap as shown below.
Replacing the connector cover

Monitor (

4, 13, 17) MENU button ( 53)
Connector cover ( 41, 42, 47)
Zoom button ( 14). See also: (thumbnail playback): 37 (playback zoom): 39 (help): 55

Multi selector (

Memory card slot cover ( 11) Flash lamp ( 18) Tripod socket (delete) button 17, 37, 38)
Mode selector 13, 23, 33) (playback) button ( 17, 37) (auto) mode ( 13)

(scene) mode (

(movie) mode ( 33)
The Multi Selector Press the multi selector up, down, left, or right to highlight pictures and menu items, then press the center to select. The buttons on the multi selector are also used to:
( : Display self-timer menu ( 20) : Display macro closeup mode menu ( 22) : Display flash mode menu 18) : Perform D-lighting ( : Transfer picture ( 40) 42)

The Monitor

The following indicators may appear in the monitor during shooting and playback (actual display varies with current camera settings):

Shooting

Shooting mode. 13, 23, 33 Exposure lock.. 31 Macro close-up mode.. 22 Zoom indicator. 14, 22 Focus indicator.. 15 Flash mode. 18 Camera shake icon. 16 Internal memory/Memory card indicator.. 11 Battery level indicator.. 13 Sensitivity.. 16 Date not set/Time zone indicator.. 57, 81
12 Self-timer indicator. Date imprint/Date counter. Number of exposures 19 20
remaining. 13, 91 Movie length. 33, 91 Focus area. 15 Image mode.. 65 Exposure compensation. 68 Color options.. 70 White balance.. 66 Best Shot Selector (BSS)/Continuous shooting mode. 68, 69

Playback

OK : START 8

15.05.2006 12:00

100NIKON 9999.JPG

9999 / 9999

1 Current folder. File number and type. Internal memory/Memory card

indicator.. 11

4 Battery level indicator.. D-Lighting guide.. Current frame number/Total
number of frames/ Movie length.. 34, 37
D-Lighting icon. 40 Movie playback indicator. 34 Image mode.. 65 Movie indicator. 34 Protect icon.. 73 Print-order icon. 50 Transfer icon.. 43, 73 Time of recording.. 9 Date of recording.. 9
First Steps Inserting Batteries
The camera uses two AA batteries. Depending on the country or region of purchase, the package will include either alkaline batteries or an MH-71 battery charger with rechargeable EN-MH1 batteries.

Turn the camera off. Confirm that the power-on lamp is off before inserting or removing memory cards. 11
Insert the memory card. Open the memory card slot cover. Slide the memory card in as shown until it clicks into place. Close the memory card slot cover.
Turn the camera on. If no message is displayed in the monitor, the camera is ready for use. If the message shown at right is displayed, the memory card must be formatted before use. CARD IS NOT FORMATTED Note that formatting permanently deletes all pictures and other data on the memory Format card. Be sure to make copies of any pictures No OK Confirm you wish to keep before beginning formatting. To begin formatting, press the multi selector up to highlight Format and then press the center of the multi selector. Do not turn the camera off or remove the batteries or memory card until formatting is complete.
Removing Memory Cards Before removing memory cards, turn the camera off and confirm that the power-on lamp is off. Open the memory card slot cover and press the card in to partially eject the card. The card can then be removed by hand. The Write Protect Switch SD cards are equipped with a write protect switch. Pictures cannot be recorded or deleted and the memory card cannot be formatted when this switch is in the Lock position. Write Protect Switch
Basic Photography and Playback

(Auto) Mode

This section describes how to take pictures in (auto) mode, an automatic, point-and-shoot mode recommended for first-time users of digital cameras.

Step1 Select

(auto) mode
Slide the mode selector to

Step2 Turn the camera on

Turn the camera on. The power-on lamp will light and the monitor will turn on. To turn the camera off at any time, press the power switch again. Do not remove the batteries while the power-on lamp is lit. Check indicators in the monitor. Check the battery level and number of exposures remaining.

Display

NO ICON Battery Level

Battery level

Batteries fully charged. Low battery; ready spare batteries.
WARNING!! BATTERY EXHAUSTED
Batteries exhausted. Replace batteries.
Number of exposures remaining ( 91)

Step3 Frame a picture

Ready the camera. Hold the camera steadily in both hands, keeping your fingers and other objects away from the lens and flash.
Frame the picture. Frame the subject near the center of the monitor.
Zoom Use the button to frame the subject in the monitor. Press to zoom out, increasing the area visible in the frame. Press to zoom in up to 3 so that the subject fills a larger area of the frame. Pressing for about two seconds at 3 zoom will trigger digital zoom, allowing the subject to be magnified up to 4 more, for a total of magnification of 12. Note that unlike optical zoom, digital zoom does not increase the amount of detail visible in the picture. Instead, details visible at 3 zoom are simply enlarged, producing a slightly grainy image. Zoom out Zoom in Zoom indicator turns yellow when digital zoom is in effect.

10.10.2006 15:30 100NIKON 0005.JPG
Deleting Unwanted Pictures
To delete the picture currently displayed in the monitor, press the button. A confirmation dialog will be displayed; highlight one of the following options and press the center of the multi selector: No: Exit without deleting the picture. Yes: Delete the picture.

Erase 1 image (s)?

Using the Flash
The flash has a range of 0.43 m (1 ft. 4 in.9 ft. 10 in.) when the camera is zoomed all the way out. The range at maximum zoom is 0.41.7 m (1 ft. 4 in.5 ft. 7 in.). The following flash modes are available:

Description

Auto Flash fires automatically when lighting is poor. 19). Auto with red- Reduces red eye in portraits ( eye reduction Off Fill flash Slow sync
Flash will not fire even when lighting is poor. Flash fires whenever picture is taken. Use to fill-in (illuminate) shadows and back-lit subjects. Flash illuminates main subject; slow shutter speeds are used to capture background at night or under dim light.

Flash OK :Confirm

Press the flash mode button.
Highlight the desired mode and press the center of the multi selector. The selected flash mode icon appears at the top of the monitor. If AUTO is selected, the icon extinguishes automatically after five seconds. The selection made in (auto) mode is stored in memory even after the camera has been turned off.
The setting will be canceled if five seconds pass without pressing the center of the multi selector.
Red-eye Reduction The camera uses advanced red-eye reduction. Low-intensity light pulses cause the subjects pupils to contract before the flash fires. If the camera still detects red-eye, it will process the image to further reduce its effects (In-Camera Red-eye Fix; note that this may slightly increase the time needed to record the picture). This mode is not recommended where a quick shutter response is required; if the desired results are not achieved, try again at a different flash setting. The Flash Lamp The flash lamp shows flash status when the shutterrelease button is pressed halfway. On: Flash will fire when picture is taken. Blinks: Flash charging. Wait a few seconds and try again. Off: Flash will not fire when picture is taken.
Taking Pictures with the Self-Timer
The camera is equipped with a ten-second timer for self-portraits. When using the self-timer, mount the camera on a tripod (recommended) or rest it on a flat, level surface.
OFF Self-timer OK :Confirm OFF ON
Press the self-timer button.

Assist modes

Select

FACE-PRIORITY AF

Press the MENU button.

Highlight , , , or and press the center of the multi selector. The framing assist menu is displayed.
Highlight the desired composition type and press the center of the multi selector. The framing guide appears.
Position the subject in the framing guide and take the picture.
Portrait Assist Use for smooth, natural-looking portraits.
FACE-PRIORITY AF PORTRAIT
Camera automatically focuses on face of portrait subject ( 26). Camera focuses on subject in center of monitor. No guides are displayed.
PORTRAIT RIGHT PORTRAIT CLOSE-UP PORTRAIT COUPLE
PORTRAIT LEFT Compose shot with subject in left or right half of frame. Camera focuses on subject in framing guide. Compose shot with subjects face in top half of frame. Camera focuses on face area in framing guide. Compose shot with two subjects side-by-side. Camera focuses on subject in framing guide.
PORTRAIT FIG- Compose shot in tall orientation. Camera focuses on subject URE in framing guide.
Taking Pictures with Face-Priority AF
Highlight and press center of multi selector. icon blinks in monitor.
Frame picture using size of icon as a guide. Closest portrait subject is indicated by double yellow border.*
* If camera detects multiple faces, closest subject will be indicated by double border, other subjects by single border. Camera will focus on closest subject. If camera can no longer detect subject (for example, because subject has looked away from camera), borders will no longer be displayed and icon will blink in monitor.
Set focus and exposure. Double border turns green.

Take the picture.

Face-Priority AF The cameras ability to detect faces depends on a variety of factors, including whether or not the subject is facing the camera. The camera may be unable to detect faces that are hidden by sunglasses or other obstructions or that take up too much or too little of the frame. If no face is detected when the shutter-release button is pressed halfway, the camera will focus on the center of the frame. The camera will focus continuously until a face is detected. If the double border flashes yellow when the shutter-release button is pressed halfway, the camera is unable to focus; release the button and try again.
Landscape Assist Use for vivid landscapes or portraits that feature a landmark in the background. Pictures taken at slow shutter speeds are processed to reduce noise ( 16).
LANDSCAPE SCENIC VIEW ARCHITECTURE
Camera focuses at infinity; no guides are displayed in monitor.* Use to photograph distant skylines. Camera focuses at infinity.* Use to photograph buildings. Camera focuses at infinity.*

Pressing the ( ) button in full-frame playback displays pictures in contact sheets of four or nine thumbnail images. The following operations can be performed while thumbnails are displayed:

Highlight picture

Delete highlighted picture Exit to full-frame playback View playback menu Exit to shooting mode

MENU 71

Change number of pictures displayed

Deleting Pictures

Pressing the button in full-frame or thumbnail playback displays the dialog shown at right. Highlight Yes and press the center of the multi selector to delete the selected picture.
Viewing Pictures in Internal Memory Pictures in internal memory can only be viewed if no memory card is inserted.
Taking a Closer Look: Playback Zoom
During full-frame playback ( 37), the ( ) button can be used to zoom in on pictures to a maximum of 10. The following operations can be performed while pictures are zoomed in:
:SCROLL SCROLL :ZOOM ZOOM
Zoom in Zoom out View other areas of picture

( ( ) )

Delete picture View playback menu Exit to full-frame playback Create cropped copy (see below) Exit to shooting mode Shutter-release button MENU

Creating a Cropped Copy

If a icon is displayed in the monitor during playback zoom, the portion of the picture currently visible in the monitor can be saved to a separate file. Pressing the shutter-release button displays the dialog shown at right; highlight Yes and press the center of the multi selector to create a new picture containing only the area visible in the monitor.
Save this image as displayed?
Depending on the size of the original and the zoom ratio at the time the copy was created, copies will be 2272 1704, 2048 1536, 1600 1200, 1280 860, 1024 768, 640 480, 320 240, or pixels in size. Cropped pictures are stored in JPEG format (compression ratio 1:8) as separate pictures from the original pictures.
Cropped Copies Cropped copies may not display or transfer correctly when viewed on other models of 92). Nikon camera. See the Appendix for more information on cropped copies (
Enhancing Contrast: D-Lighting

Print selection Print all images DPOF printing
PictBridge Print selection Print all images DPOF printing Paper size Cancel OK Confirm
Print selected pictures (see below). Print one copy of all images. Press multi selector up or down to highlight option, press center to select: Start print: Print current print order ( 52). Confirm: View print order. Press center of multi selector to start printing, or press MENU to exit. Cancel: Exit without printing. Choose page size ( 49).
DPOF PRINTING 005 prints Start print Confirm Cancel OK Confirm
Printing Selected Pictures Choosing Print selection displays the menu shown in Step 1.

PRINT SELECTION

PRINT SELECTION 1

10. 10. : 00

+ -[ 4/ 5]

MENU Back OK Confirm

Scroll through pictures. Current picture shown at center of display.
Select current picture and set number of prints to 1. Selected pictures marked by icon.
PRINT SELECTION 3 10. 10. : 00 + -[ 4/ 5]

005 PRINTS

Choose number of prints (up to 9). To deselect picture, press multi selector down when number of prints is 1. Repeat steps 12 to select additional pictures.
Confirm selection (to return to step 1 and change selection, press MENU button).

Printing 001/003

OK Cancel
Start printing. PictBridge menu is displayed when printing is complete.*
* Press center of multi selector again to cancel before printing is complete.
Creating a DPOF Print Order: Print Set
The Print set option in the playback menu ( 71) is used to create digital print orders for printing on DPOF-compatible devices ( 90). Selecting Print set displays the menu shown in Step 1.
PRINT SET Print selected Delete print set
10. 10. : 00 MENU Exit OK Confirm
Highlight Print selected.
Display selection dialog.
* To delete current print order, highlight Delete print set and press center of multi selector.
Select pictures as described in Steps 14 of Printing Selected Pictures ( 50).
PRINT SELECTION Done Date Info MENU Exit OK Confirm
Press multi selector up or down to highlight option, press center to select.
Date: Print date of recording on all pictures in print order. Info: Print shutter speed and aperture on all pictures in print order. Done: Complete print order and exit.
Print Set Date and Info are reset each time the menu shown Step 4 is displayed. Info is not printed when the camera is connected directly to the printer.
Menus are available for shooting, playback, movies, and camera setup. Press the MENU button to display the menus; press again to return to shooting or playback. To display the shooting menu:
SHOOTING MENU 1/2 SET-UP Image mode White balance Exp. +/Continuous MENU Exit OK Confirm ? Help
To display the playback menu:
PLAYBACK MENU 1/2 SET-UP Print set Slide show Delete Protect MENU Exit OK Confirm ? Help

Choose how menus are displayed. Select screen displayed when camera is turned on. Set clock and choose home and destination time zones. Adjust monitor brightness and choose display style. Imprint date or birthday counter on pictures. Adjust sound and volume settings. Turn blur warning on or off. Choose when monitor will turn off to save power. Format internal memory or memory card. Choose language for camera menus and messages. Adjust settings for connection to television or computer. Restore camera settings to default values. Choose type of battery inserted in camera. Display camera firmware version. 64 64
Default Settings See the Appendix for information on setup menu defaults (
Choose how menus are displayed.
SET-UP Shooting menu Menus Welcome screen Date Monitor settings MENU Exit OK Confirm ? 1/3 WELCOME IMAGE SELECT MENUS Text Icons
MENU Exit OK Confirm MENU Exit OK Confirm ? Help

Welcome Screen

Choose the welcome screen displayed when the camera is turned on. To select a picture from internal memory or the memory card to use as a welcome screen, choose Select an image (the image used in the welcome screen is a copy of the selected image and does not change when the original is deleted or the memory card is removed).
WELCOME IMAGE SELECT Disable welcome Nikon Animation Select an image

COOLPIX

Set the camera clock and choose home and travel destination time zones.

Date Time zone

DATE 10.10.2006 15:30
Set camera clock to current date and time ( 9). Choose home ( ) or travel destination ( ) time zone ( 93); turn daylight saving time on or off.
Date Time zone MENU Exit OK Confirm --

To switch between home (

) and travel destination (

) time zones:

To switch between home ( ) and travel destination ( ) time zones: Highlight or and press the center of the multi selector. The clock is set to the time in the selected time zone. The current selection is indicated by a dot.
TIME ZONE 10.10.2006 15:30 London, Casablanca DaylightSaving OK London, Casablanca DaylightSaving MENU Exit OK Confirm
* To change the home or travel destination time zone, highlight or , then press the multi selector right. Choose a time zone and daylight saving time as described in Basic Setup ( 9).
The Clock Battery The clock battery charges when the main battery is installed or an AC adapter is connected, and can provide several days of backup power after about 10 hours of charging.

Alkaline COOLPIX (NiMH) Lithium
BATTERY TYPE Alkaline COOLPIX(NiMH) Lithium MENU Exit OK Confirm
Alkaline EN-MH1, oxyride Lithium

Firmware Version

View the current camera firmware version.

COOLPIX L4 Ver.X.X

Shooting Options: The Shooting Menu
With the exception of Image mode, the following options are available only in (auto) mode. See Menus for information on viewing and using the shooting menu ( 53).
SET-UP Image mode White balance* Exposure compensation Continuous* BSS* Color Options*
Display the setup menu. Choose image size and quality. Match white balance to the light source. Exposure can be compensated for in the case of a subject that is bright, dark or with strong contrast. Take pictures one at a time or in a sequence. Use BSS to choose the sharpest picture. Control color. 66 68
* See the Appendix for information on restrictions that may apply to these settings ( 92). Shooting Menu Settings See the Appendix for information on shooting menu defaults ( 88).

Image Mode

Pictures taken with a digital camera are recorded as image files. The size of the files, and with it the number of images that can be recorded ( 91), depends on the size and quality of the images (image mode). Before shooting, choose an image mode according to how you plan to use the picture.

High(2272) Normal(2272)

IMAGE MODE High(2272 ) Normal(2272) Normal(1600) PC screen(1024) TV screen(640) MENU Exit OK Confirm

2272 1704

Pictures are compressed less than Normal, producing higher quality images. Best choice in most situations.
Normal(1600) PC screen (1024) TV screen(640)

640 480

Smaller size allows more pictures to be stored. Suited to computer display. Suited to e-mail or television viewing.
The current setting is shown by an icon in the display.

White Balance

Process pictures to produce natural colors under different types of light.
1/2 WHITE BALANCE Auto White bal. preset Daylight Incandescent Fluorescent MENU Exit OK Confirm

Natural color.

Vivid color Vivid, photoprint colors. Black-and- Take pictures in black-and-white. white Sepia Record pictures in sepia. Cyanotype Take pictures in blue-tinted monochrome.
The effects of the selected option can be previewed in the monitor. At settings other than Standard Color, the current setting is also indicated by an icon in the display.
Playback Options: The Playback Menu
The playback menu contains the following options. See Menus for information on viewing and using the playback menu ( 53).
SET-UP Print set Slide show Delete Protect Transfer marking Small pic. Copy
Display the setup menu. Select pictures for printing. View pictures in an automatic slide show. Delete all or selected pictures. Protect selected pictures from accidental deletion. Change transfer marking for existing pictures. Create a small copy of the current picture. Copy files between memory card and internal memory. 73 73

Print Set

Create digital print orders for PictBridge printers and devices that support DPOF ( 52).

Slide Show

Play pictures back in an automated slide show, with about three seconds between each picture.

SLIDE SHOW Pause OK

Start Loop MENU Exit OK Confirm

Highlight Start.

Begin slide show.
Movies Movies appear in slide shows as still images of the first frame. Loop Select Loop to repeat the slide show until the center of the multi selector is pressed. Note that the monitor will turn off automatically if no operations are performed for 30 minutes.
To skip to the next picture, press the multi selector to the right, or press the multi selector to the left to return to the previous picture. Hold the multi selector down to fast forward or rewind. To pause the slide show, press the center of the multi selector. The menu shown below will be displayed. When the slide show ends or is paused, the menu Pause shown at right will be displayed. Highlight Restart and press the center of the multi selector to restart the show, or select End to end the show and return to the playback menu. End OK :Confirm

Restart

Delete
Delete all or selected pictures. Note that pictures marked with a icon are protected and cannot be deleted.
DELETE Erase selected images Erase all images

Protect

Protect selected pictures from accidental deletion. Protected pictures are marked with a icon (see right, below) and cannot be deleted using the button or the Delete option in the playback menu. Note, however, that formatting will delete protected files ( 63).

Date imprint not available Date not appearing on pictures File cannot be played back Cannot zoom in on picture
Cannot use crop, D-lighting, or small picture options
Cannot mark pictures for transfer
Transfer marking does not appear
PictureProject does not start when camera is connected
Camera is off. Batteries are exhausted. USB cable is not correctly connected. Appropriate USB option is not selected. Camera is not registered in Device Manager (Windows only). See the PictureProject reference manual (on CD) for further information.
Pictures not displayed on television Choose correct video mode. Format memory > Format option not available Batteries are low.

41 13, 63

Electronically Controlled Cameras In extremely rare instances, unusual characters may appear in the monitor and the camera may stop functioning. In most cases, this phenomenon is caused by a strong external static charge. Turn the camera off, remove and replace the batteries, and then turn the camera on again. In the event of continued malfunction, contact your retailer or Nikon representative. Note that disconnecting the power source as described above may result in the loss of any data not recorded to internal memory or the memory card at the time the problem occurred. Data already recorded will not be affected.

Appendix

Defaults ( 54, 56, 65) Choosing Reset for Reset all restores the following default settings:
mode Flash mode ( Self timer ( 18) 20)

Default

Auto OFF OFF
Scene mode Portrait assist ( Landscape assist ( 27) Sports assist ( 27) 25)
Macro close-up ( 22) Setup menu Welcome screen ( 57) Photo info ( Brightness ( Date imprint ( Button sound ( Shutter sound ( 61) Start-up sound ( 61) Blur warning ( Auto off ( 62) 62) 62)
Night portrait assist ( 28) Animation Show info 3 Off Shooting menu Image mode ( White balance ( 66) 65) Normal (2272) Auto

59) 59) 60)

Exposure Compensa- 0 tion ( 68) Continuous ( BSS ( 69) 68) Single Off Standard color

61) ON ON ON On 1m Off

Color Options ( 70) Movie menu Movie options ( 35) Auto-focus mode ( 36) PictBridge Paper size ( 49)

Small size 320 Single AF

Specifications

Nikon COOLPIX L4 Digital Camera
Type Effective pixels CCD Image size (pixels) Lens Focal length f/-number Construction Digital zoom Autofocus (AF) Focus range (from lens) Focus-area Monitor Approximate frame coverage Storage Media File system File formats Exposure Metering Exposure control Range
Matrix, center-weighted Programmed auto exposure with exposure compensation (2.0+2.0EV in steps of 1/3EV) W: +1.0+17.1EV, T: +2.7+18.8EV Internal memory (approx.10 MB); SD (Secure Digital) memory cards DCF, Exif 2.2, and DPOF compliant Compressed: JPEG-baseline-compliant Movies: QuickTime
Compact digital camera 4.0 million 1/2.5-in. CCD; total pixels: 4.23 million

(2272, 2272) (1600)

(1024) (640)
Zoom-Nikkor with 3 optical zoom F=6.318.9 mm (35-mm [135] camera-format equivalent: 38114 mm) f/2.8f/4.elements in 6 groups Up to 4 (35-mm [135] camera-format equivalent: 456 mm) Contrast-detect AF Approximately 30 cm (1 ft.) Macro close-up mode: Approximately 4 cm (1.6 in.) (W) Center 2.0 in., 115,000-dot, TFT LCD monitor with brightness adjustment Shooting mode: 97% horizontal and 97% vertical Playback: 100% horizontal and 100% vertical
Shutter Speed Aperture Range ISO sensitivity Self-timer Built-in flash
Mechanical and charge-coupled electronic shutter 41/3000 s Magnetically controlled f/2.8 and f/4.9 (W) Approximately equivalent to ISO 50 (auto gain to ISO 200) Approximately 10 s
Range (approx.) W: 0.43.0 m/1 ft. 4 in.9 ft. 10 in.
T: 0.41.7 m/1 ft. 4 in.5 ft. 7 in.
Sync method I/O terminals Interface Video output Supported languages Power sources
Sensor flash system Video out/digital IO (USB) USB Can be selected from NTSC and PAL Chinese (Simplified and Traditional), Dutch, English, French, German, Italian, Japanese, Korean, Russian, Spanish, Swedish
Two AA alkaline, oxyride, or lithium batteries Two rechargeable EN-MH1 NiMH batteries EH-65A AC adapter kit
Battery life Approximate dimensions Approximate weight
Approximately 250 shots with alkaline, 600 shots with lithium, or 450 shots with EN-MH1 batteries* 86.5 60.5 34.5 mm/3.4 2.4 1.4 in. (W H D) 115 g (4.1 oz.) without battery or memory card

Operating environment Temperature Humidity
0+40C (+32104F) Less than 85% (no condensation)
* Based on Camera and Imaging Products Association (CIPA) standard for measuring life of camera batteries. Measured at 25C (77F); zoom adjusted with each shot, flash fired with every other shot, image mode set to NORMAL.
Unless otherwise stated, all figures are for a camera with fully-charged EN-MH1 batteries operated at an ambient temperature of 25C (77F).
MH-71 Battery Charger (Option; Supplied with Camera in Some Regions) Rated input Charging output Applicable batteries Discharge time Charging time Operating temperature Approximate dimensions Length of cord Approximate weight
AC 100240 V, 0.110.065 A, 50/60 Hz DC 1.2 V/2000 mA Nikon EN-MH1 rechargeable Ni-MH batteries Approximately 5 hours Approximately 2.5 hours 535C (+4195F) 30 mm/2.8 3.1 1.2 in. (W H D) Approximately 1800 mm (70.9 in.) 110 g (3.9 oz.), excluding power cable
Specifications Nikon will not be held liable for any errors this manual may contain. The appearance of this product and its specifications are subject to change without notice. EN-MH1 NiMH Batteries The capacity of NiMH batteries will drop if they are charged repeatedly without first being fully discharged. The batteries will return to normal once fully discharged. Note that NiMH batteries gradually lose their charge when not in use; refresh the batteries immediately before use.

Symbols

Zoom button (help), 3, 55 (thumbnail), 3, 38 (playback zoom), 3, 39 (auto) mode, 3, 13 (playback) button, 3, 17, 37 Flash mode, 3, 18 Macro close-up mode, 3, 22 (delete) button, 3, 17, 38 Transfer picture, 3, 43 (movie) mode, 3, 33 Self-timer, 3, 20 D-lighting, 3, 40 (scene) mode, 3, 23 button, 3, 53 Blur, 16, 62, 81, 85 Blur warning, 62 Brightness, 59 BSS (best shot selector), 69 Button sound, 61
Face-priority AF, 26 Fill flash, 18 Firmware, 64 Firmware version, 64 Flash, 18, 88, 90 Flash lamp, 3, 19 Flash mode, 18, 90 Focus, 4, 15, 36 area, 15 indicator, 15 lock, 16 mode, 36 Format, 12, 63 Format card, 63 Format memory, 63 Framing, 14, 24 assist, 24 guide, 24 Framing assist, 24
Camera strap, 2 Clock, 57 Close-up, 22 Color options, 70 Computer, 41, 42, 64 Connector cover, 3, 41, 42, 47 Copy, 39, 40, 75 Cropping picture, 39 Cyanotype, 70
Date, 57 Date and time printing, 60 Date counter, 60 Date imprint, 52, 60 DCF (Design rule for Camera File System), 89 Default setting, 88 Delete, 73 Deleting picture, 17, 38, 73 D-lighting, 40 DPOF (Digital Print Order Format), 50, 89
Accessories, 76 AE-L, 31 AF (auto-focus) indicator, 15 Assist modes, 24 Auto off, 62 Auto-focus mode, 36

Voice of the Customers: Mining Online Customer Reviews for Product Feature-based Ranking
Ramanathan Narayanan Alok Choudhary Electrical Engineering and Computer Science Department Northwestern University 2145 Sheridan road, Evanston, IL, 60208, USA Abstract
Increasingly large numbers of customers are choosing online shopping because of its convenience, reliability, and cost. As the number of products being sold online increases, it is becoming increasingly difcult for customers to make purchasing decisions based on only pictures and short product descriptions. On the other hand, customer reviews, particularly the text describing the features, comparisons and experiences of using a particular product provide a rich source of information to compare products and make purchasing decisions. Online retailers like Amazon.com1 allow customers to add reviews of products they have purchased. These reviews have become a diverse and reliable source to aid other customers. Traditionally, many customers have used expert rankings which rate limited a number of products. Existing automated ranking mechanisms typically rank products based on their overall quality. However, a product usually has multiple product features, each of which plays a different role. Different customers may be interested in different features of a product, and their preferences may vary accordingly. In this paper, we present a feature-based product ranking technique that mines thousands of customer reviews. We rst identify product features within a product category and analyze their frequencies and relative usage. For each feature, we identify subjective and comparative sentences in reviews. We then assign sentiment orientations to these sentences. By using the information obtained from customer reviews, we model the relationships among products by constructing a weighted and directed graph. We then mine this graph to determine the relative quality of products. Experiments on Digital Camera and Television reviews from real-world data on Amazon.com are presented to demonstrate the results of the proposed techniques.

Kunpeng Zhang

1 Introduction
The rapid proliferation of Internet connectivity has led to increasingly large volumes of electronic commerce. A study conducted by Forrester Research[1] predicted that e-commerce and retail sales in the US market during 2008 were expected to reach $204 billion, an increase of 17% over the previous year. More customers are turning towards online shopping because it is convenient, reliable, and fast. It has become extremely difcult for customers to make their purchasing decisions based only on images and (often biased) product descriptions provided by the seller. Online retailers aim to provide consumers a comprehensive shopping experience by allowing them to choose products based on their specic needs like price, manufacturer, and other attributes. They also allow customers to add reviews of products they have bought. Customer reviews of a product are generally considered more honest, unbiased and comprehensive than descriptions provided by the seller. Furthermore, reviews written by other customers describe the usage experience and perspective of (non-expert) customers with similar needs. A study by comScore and Kelsey group[2] showed that online customer reviews have signicant impact on prospective buyers. However, as the number of customer reviews available increases, it is almost impossible for a single user to read all reviews and comprehend them to make informed decisions. Therefore, mining these reviews to extract useful information efciently is an important and challenging problem. The abundance of customer reviews available has led to a number of scholars doing valuable and interesting research related to mining and summarizing customer reviews[3, 4, 5, 6, 7, 8]. There has also been considerable work on sentiment analysis of sentences in reviews, as well as sentiment orientation of a review as a whole[9]. There has been relatively little work on ranking products 1

1 http://www.amazon.com

negative sentiments about the product being reviewed. Comparative sentences contain information comparing competing products in terms of features, price, reliability etc. Fig. 1 depicts a typical customer review highlighting the different kinds of sentences. After developing techniques to identify such sentences, we build a weighted and directed product graph(feature-based) that captures the sentiments expressed by customers in reviews. We perform a ranking of products based on the graph. Particularly, the main contributions in this work are: NLP and dynamic programming techniques to identify subjective/comparative sentences(product feature-based) in reviews and determine their sentiment orientations. Sentence classication techniques to build a weighted and directed graph which reects the inherent quality of products in terms of their features. A ranking algorithm that uses this massive graph to produce a ranking list of products based on each considered product feature. That is, the end result of the algorithm is a ranking list that a potential customer can use to determine the best products based on customers importance of one or more product features. The reminder of this paper is organized as follows. Section 2 contains a summary of related work. We present our techniques in Section 3. Section 4 contains the details of datasets and experimental results followed by the conclusions in Section 5.
Figure 1: A Customer Review from Amazon.com. Product Features and Subjective/Comparative Sentences are Highlighted.
automatically based on customer reviews. Our work is based on the our proposed ranking scheme, where products are ranked based on their overall quality. A product feature is dened as an attribute of a product that is of interest to customers. Even though an overall ranking is an important measure, different product features are important to different customers based on their usage patterns and requirements. And different products may be designed and rank differently based on the feature of interest. For instance, a digital camera that is ranked highly overall may have less-than-stellar battery life. Thus, both overall ranking and more detailed product feature based ranking are important. In this paper, we propose an algorithm that uses customer review text, mines tens of thousands of reviews, and provides ranking of products based on product features. For each product category, product features are dened and extracted in the data preprocessing step. Note that for most products, there is a standard set of features which are considered important and normally are provided with product descriptions. We then label each sentence with the product features described in it. We then identify four different types of sentences in customer reviews that are useful in determining a products rank: positive subjective, negative subjective, positive comparative, negative comparative. Subjective sentences are those sentences in which the reviewer expresses positive or 2

2 http://www.consumerreports.org/
In our review dataset containing 1516001 sentences, we observed that around 16% of sentences describe one or more of these features. To tag each sentence with features, we use a simple strategy: if the sentence contains one of the words/phrases in the synonym set for a product feature, we mark it as describing the feature. Since we have dened these features manually, we describe them in greater detail. For Digital Cameras, ash is an important feature for indoor and low-light photography. Battery life is a sought-after feature that details the kind of batteries used. Focus talks about auto-focus or manual focus capabilities. Lens is a critical factor for professional photographers purchasing high-end cameras. The Optical feature encompasses digital zoom and optical zoom. LCD represents the digital display/screen that lets a user see how a photo will look like. Resolution refers to the sharpness, or detail, of a picture. Burst is used to describe the rapid re and continuous shooting capabilities. Memory determines the number of pictures that can be taken and Compression determines how le size of a photo is shrunk. For the Television segment, the Sound feature is useful for users interested in audio quality (some TVs come with an extra set of speakers to create surround sound). Reectivity/Anti-glare is important for the viewing experience. Size represents the size, height, weight of a television screen. Connections means the number and type of input ports available for hooking up devices to the television. The richness/quality of the images displayed are described in Picture quality. Users are also interested in the remote control device available with the television. Resolution refers to the number of pixels or lines displayed on the screen. Adjustment is the ability/mode that expands or compresses an image to ll the screen better. Picture-in-picture(PIP) feature allows a user to watch two channels at once. FilmMode/CineMotion improves the movie-watching experience, which may be important to some users.

3.2 Sentence Labeling

Customers express their opinions about products in multiple ways. We identify two kinds of sentences that are useful while ranking products: Subjective Sentences and Comparative Sentences. Before doing this task, we use MxTerminator[11] to split reviews into sentences because a typical customer review comprises of several sentences. We formally dene the different types of sentences below: Denition 3.1. Subjective Sentence(SS) A sentence expressing direct praise or deprecation about a product. Ex. This camera has excellent shutter speed. Denition 3.2. Comparative Sentence(CS) A sentence which indirectly express an opinion by performing a 3
Table 1: Keywords Representing 10 Most Important Product Features for Digital Camera and Television Domains Digital Camera TV resolution|pixel|megapixel connection|input|output| component video|composite video|HDMI lens|wide angle|normal range adjustment|stretch|zoom|expand|compress optical|zoom|optical zoom|digital zoom lm-mode|frame|theatrical|3:2|pull-down| motion compensation|CineMotion memory|megabytes|MB pip|picture-in-picture|dual-tuner|pop| picture-outside-picture|two-tuner burst|continuous|shutter|recovery|motion|sport resolution|1080p|1080i|720p battery|batteries|power screen|anti-glare|reectivity|burn-in|shiny| screensaver|pixel-shift focus|exposure|manual|iso picture|image|picture quality|image quality LCD|screen sound|sound quality|speaker|stereo|audio compression|compress|jpeg size|height|width|depth|weight|inch ash|light remoter|remote|gear|universal comparison between two products. Ex. I think the coolpix is better than the canon sd1200. Denition 3.3. Product Comparative Sentence(P CS) A PCS is a comparative sentence and contains at least one product name. Ex. This TV has much better sound quality when compared to the sony bravia. propose a more rened technique to nd comparative sentences specically related to product comparisons in our previous work[21]. We use a dynamic programming technique (longest common subsequence) to identify product-product comparison pairs in a comparative sentence. We use only comparative sentences which contain at least one product name which is different from the product the sentence is describing while building our ranking model. In [21], we have shown that we get a precision of 82% and a recall of 80% approximately.

Identifying Comparative Sentences
There has been some earlier research[16, 17] regarding identication of comparative sentences in text. These techniques use keyword comparison(KW contains 126 words, some of which are explicit(outperform, exceed, compare, superior, etc.) and others are implicit(prefer, choose, like, etc.)), sentence semantics, and sentence structure to identify comparative sentences. To identify part-of-speech tags, CRFTagger[12], a java-based conditional random eld part-of-speech (POS) tagger for English is employed. We build on these techniques and use the following rules for identifying comparative sentences: Check if the sentence contains any comparative keywords in KW ; Recognize any words with POS tags { JJR(comparative adjective), RBR(comparative adverb), JJS(superlative adjective), RBS(superlative adverb) }; Scan if any predened structural patterns are present in the sentence (as <word> as, the same as, similar to, etc.). Note that not all sentences satisfying these rules are comparative sentences in terms of product comparison. For example, the sentence I bought this camera for my son because he got a higher grade in his second statistical exam. does not show any comparative meanings or implications over other camera products. Therefore, we 4
3.4 Identifying Sentence Sentiment Orientation
In this section, we describe how we assign sentiment orientations for a sentence. We only consider positive and negative sentiments in this work. Unfortunately, dictionaries and other sources like WordNet[13] do not include sentiment orientation information for each word. Some researchers[18] have used supervised learning algorithms to infer the sentiment orientation of adjectives and adverbs from constraints on conjunctions. [9] contains a summary of existing sentiment analysis techniques. In this paper, we use a simple yet powerful method by utilizing a positive word set(POS) and a negative word set(NEG) developed in the MQPA project[19]. We also add some words of our own. At the end of this process, we get a list of 1974 words for the positive set and 4605 words for the negative set. We use a simple technique to identify the orientation of a sentence using these words. If the sentence contains a word that is in the positive word set, we label this sentence with a positive tag. Negative sentiment words are handled similarly. However, many customers do not like to express their opinions by using assertive sentences but using some negations in their reviews. In this case, the orientation should be switched. We constructed a set of 28 negation

words manually. It should be mentioned that we determine the sentence orientation for comparative sentences as well using the same list of sentiment words. A precision of 82% and a recall of 86% approximately can be achieved and shown in [21].
Table 2: Product Ranking Results for Gf in Fig. 2 Rank Vertex ID Score 1 B 0.D 0.C 0.A 0.052781 weight is also crucial to the ranking, in addition to the edge weights. The idea can be formalized using the equation below: pRank(P ) = [(1 d) + d pRank(Pi ) Ce (Pi )] Cv (P ), where
Constructing the Product Graph
We use the subjective and comparative sentences found to construct a directed and weighted graph for each product feature that can be mined to reveal the relative quality of products. The product graph for feature f is dened as follows: Gf = (V, E) where V is a set of nodes, V = {pi | each node represents a product, 0 < i < n}, E is a set of node pairs, called arcs or directed edges. An arc e = (pi , pj ) is considered to be directed from pi to pj. E = {ek = (pi , pj ), | Wei is the weight of the edge ei , 0 < i, j < n, 0 < k < m}, where n is the number of products, m is the number of edges. Consider a comparative sentence in the reviews for product Pi describing feature f. If this sentence compares product Pi with product Pj , we add a directed edge from Pj to Pi. The second step is to assign a weight to this edge. A comparative sentence occurring in the reviews for product Pi and comparing it with product Pj is considered a positive comparative(P C(Pi , Pj )) if it implies that Pi is better than Pj. If it implies that Pi is worse that Pj , it is considered a negative comparative(N C(Pi , Pj )). For each edge(Pj , Pi ), we count the number of positive (P C) and negative (N C) comparative sentences associated with the pair (Pi , Pj ) respectively. We assign the ratio P C/N C as the weight of the edge linking Pj to Pi. The weight of a node is used to represent the inherent quality of a product. For a node Pi , we use the ratio of the number of positive(P S) and negative(N S) subjective sentences (P S/N S) as its weight.

1{Pi ,P }

pRank(P ) is the product ranking of product P ; pRank(Pi ) is the product ranking of product Pi and n is the number of incoming links on product P ; 1{Pi ,P } is an indicator function, s.t. 1{Pi ,P } = if there is a link from Pi to P otherwise
W (P ,P Ce (Pi ) = Pm eWei(Pi),Pj ) , where m is the number j=1 of outbound links on product Pi , Pj are the nodes pointed to from Pi and We (Pi , Pj ) is the weight of the edge (Pi , Pj ). It is the edge weight contributor to the ranking of product P ; ) Cv (P ) = Pn Wv (P,P t ,Pt ). It is the node weight t=1 Wv (P contributor to the ranking of product P.
Let us illustrate the ranking process using a simple example. We have four products(A, B, C, D) which we wish to rank according to product feature f. The numbers of positive/negative, subjective/comparative sentences labeled with feature f are listed below. P Sf (A) = 1, P Sf (B) = 2, P Sf (C) = 3, P Sf (D) = 4 N Sf (A) = 3, P Cf (B, A) = 3, P Cf (B, C) = 7 P Cf (B, D) = 3, P Cf (A, C) = 2, N Cf (B, C) = 2 Based on these sentence statistics, we build a product graph Gf (see Fig. 2). Edge weights are determined by comparative sentences, and node weights are determined by subjective sentences. Since the reviews of product C have 7 positive comparative sentences mentioning product B (and feature F ), and 2 negative comparative sentences mentioning B (and feature f ), there is an edge from C to B with weight 3.5. It must be mentioned that to prevent edges with innite length (when the number of negative comparative sentences is 0), we set the minimum value of the denominator to 1 while 5

Ranking Products

We evaluate the relative importance of each product according to feature f by using the pRank algorithm described below. The classic PageRank algorithm[20] treats all edges equally and does not take node weights into account. A node has a higher importance if it is pointed to from relatively important nodes. In pRank, we not only consider the relative importance among products, but also take the importance/quality of the product itself into account. This means that the node
Algorithm 1 pRank Rank Products for Feature f Require: Product Feature(f ), Product Category(Cat). Ensure: The ranking list of products belonging to category Cat for the feature f.

1: 2: 3: 4: 5: 6: 7: 8: 9: 10: 11: 12: 13: 14: 15: 16: 17: 18: 19: 20:
Figure 2: A Simple Ranking Example With a Product Graph Gf Having 4 Products Regarding a Specic Feature f (+ means positive and - means negative) computing edge weights. By using our algorithm, we get the ranking score for each node shown in the Table 2. The ranking order(the smaller, the product better) for this graph is B D C A. From the graph, we clearly see that A, C, D are worse than B because all of them have edges pointing to B. D has more positive subjective sentences than A, C and their comparative weights with B are approximately equal. C has a better ranking than A because (i) two sentences say A is better than C and (ii) reviews for A contain 1 positive/3 negative subjective sentences while reviews for C contain 3 positive subjective sentences. Algorithm 1 below summarizes our ranking methodology.
XML Data = Download(Cat); SENT = Get Sentence(XML Data); LSENT = Label(SENT, F ); {P S, N S, P C, N C} LSEN T ; for each sentence s {P C, N C} do Find all product comparison pairs {pi , pj } using dynamic programming; Pair {pi , pj } + P os or N eg ; end for Count P Spi , N Spi , P Cpi ,pj , N Cpi ,pj related to all products; for each product pi do for each product pj do if i == j then M atrix[i, i] = P Spi /N Spi ; else M atrix[i, j] = P Cpj ,pi /N Cpj ,pi ; end if end for end for Ranking List = Rank(M atrix[]); return Ranking List;

Experiment Results

In this section we evaluate the performance of our ranking algorithm. We conduct our experiments on customer reviews from two different product categories (Digital Camera and Televisions) from Amazon.com. Further details about the datasets and the APIs used to generate this data can be found at Amazon.com3 and BrowseNodes.com4. The Digital Camera dataset contains 83005 reviews (for 1350 products) and Television dataset contains 24495 reviews (for 760 products) collected by August, 2009. Table 3 and Table 4 show the relevant statistics for these two datasets: total number of sentences, frequency of occurrence of different product features, number of subjective and comparative sentences and their sentiment orientations. To evaluate our ranking algorithm, we rst perform product ranking based on the overall quality. To determine the overall rank of a product, we include all comparative and subjective sentences in our database while constructing

3 http://www.amazon.com/ 4 http://www.browsenodes.com/
the product graph. There is no ltering done for product features. We then mine this overall graph Goverall using the ranking algorithm described in Section 3.6. To evaluate the effectiveness of this ranking strategy, we compare our results with a ranking performed by domain experts. The results indicate that our product ranking strategy achieves signicant agreement with evaluations done by subject experts with several years of experience and insight in their respective elds. Approximately, an average overlapping probability of 62% could be achieved for different price bins for cameras and televisions. More details about this evaluation can be found in[21]. In this paper, we focus on the feature-specic ranking obtained by mining the individual product graphs generated for each product feature. Intuitively, the feature-specic ranking should not be dramatically different from the overall ranking. If we have chosen a relevant set of product features that customers are interested in, then the top-ranked products in these lists should not rank badly in the overall list. However, it is quite likely that there are signicant differences in the ranking order of these products, especially at the top. To clarify this intuition, we give the following example: If a product ranks in the top 5 products according to feature f (Ex. lens), then the probability that it ranks in the 6
Table 3: Breakdown of Subjective/Comparative Sentences(Digital Camera) Feature/Overall No. of Sentences No. of Subjective Sentences No. of Comparative Sentences Positive Negative Positive Negative Flash 514 Battery 533 Focus 720 Lens 437 Optical 338 Lcd 216 Resolution 227 Burst 189 Memory 143 Compression 29 Digital Camera 10890 Feature/Overall Sound Screen Size Connection Resolution Picture Quality Remoter Adjustment PIP Film-Mode TV Table 4: Breakdown of Subjective/Comparative Sentences(TV) No. of Sentences No. of Subjective Sentences No. of Comparative Sentences Positive Negative Positive Negative 9162
bottom 5 products overall should be very low. Similarly, if a product has high overall rank, then it should rank Table 5: The Relative Feature Fraction(RF Ff ) for Digital Camera and TV highly according to some features. Another aspect of our feature ranking methodology is the relative importance of different product features. A valid question that can be asked is: Which product features are customers looking for when making their choices? To answer this question we dene two metrics, Relative Feature Fraction(RF F ) and Importance of Feature(IF ). Digital Camera Flash Battery Focus Lens Optical LCD Resolution Burst Memory Compression RF Ff 18.18% 15.96% 15.93% 13.67% 10.77% 9.72% 5.63% 5.40% 4.06% 0.67% TV Sound Screen Size Connection Resolution Picture Quality Remoter Adjustment PIP Film-Mode RF Ff 24.76% 16.10% 12.87% 11.24% 10.98% 8.90% 8.13% 3.04% 2.15% 1.82%

Denition 4.1. Relative Feature Fraction: RF Ff = PNf *100%, where Nf is the number of sentences laf Nf Table 6: Importance of Feature(IFf ) for Digital Camera beled with feature f. and TV Denition 4.2. Importance of Feature: IFf = |XYf | IFf 100, where X = {top 10% of overall ranked Digital Camera Features IFf TV Features |X| Lens 79.9 Size 78.7 products}, and Yf = {top 10% of products according to Resolution 79.8 Film-Mode 72.3 feature f }. Optical 77.5 Picture Quality 70.7 Focus 76.3 Connection 69.1 76.3 PIP 69.1 Table 5 shows the RF F for different features in Dig- Memory Burst 75.2 Sound 67.5 ital Camera and Television categories respectively. We Lcd 74.1 Remoter 67.5 observe that customers are mentioning ash and sound Flash 72.9 Adjustment 64.3 (for Camera and TV) product features in their reviews Battery 71.6 Screen 61.2 most often. However, a higher RF F does not always in- Compression 68.4 Resolution 61.2 dicate greater importance of a feature. The importance of 7
Table 7: Top 10 Rated Digital Cameras for Each Feature and Overall Ranking Lens Resolution Optical Focus
PENTAX K100D SUPER KODAK EASYSHARE ONE CANON POWERSHOT SD500 CANON POWERSHOT SD990IS NIKON COOLPIX L3 NIKON COOLPIX 8400 NIKON COOLPIX L1 KODAK EASYSHARE C653 NIKON COOLPIX P6000 NIKON COOLPIX L4 CANON POWERSHOT SD200 NIKON COOLPIX S4 NIKON COOLPIX L3 NIKON COOLPIX S600 KODAK EASYSHARE C653 NIKON COOLPIX P90 NIKON COOLPIX L4 SONY ALPHA A700K NIKON COOLPIX P50 PENTAX OPTIO 60 CASIO EXILIM PRO KODAK EASYSHARE ONE CANON POWERSHOT SD550 CANON POWERSHOT SD990IS NIKON COOLPIX 3200 NIKON COOLPIX L3 NIKON COOLPIX S500 HP PHOTOSMART M537 NIKON COOLPIX P50 NIKON COOLPIX 7600 CASIO EXILIM PRO KODAK EASYSHARE ONE PANASONIC DMC-FX37A 10.1MP CANON POWERSHOT S60 CANON POWERSHOT SD990IS NIKON COOLPIX 3200 PENTAX OPTIO P70 NIKON COOLPIX L3 NIKON COOLPIX 8400 NIKON COOLPIX L1

Memory

KODAK EASYSHARE ONE CANON POWERSHOT S80 PANASONIC DMC-FX37A 10.1MP PENTAX OPTIO A10 CANON POWERSHOT SD990IS NIKON COOLPIX 3200 OLYMPUS SP-550UZ 7.1MP NIKON COOLPIX 4300 CANON POWERSHOT S410 NIKON COOLPIX 8400
CASIO EXILIM PRO CANON POWERSHOT SD500 PANASONIC DMC-FX37A 10.1MP CANON POWERSHOT S80 CANON POWERSHOT S60 CANON POWERSHOT SD990IS NIKON COOLPIX 3200 NIKON COOLPIX 995 CANON POWERSHOT S100 NIKON COOLPIX L3
KODAK EASYSHARE ONE PANASONIC DMC-FX37A 10.1MP CANON POWERSHOT S80 PENTAX OPTIO A10 CANON POWERSHOT SD990IS NIKON COOLPIX 3200 CANON POWERSHOT S400 CANON POWERSHOT G3 SONY DSCP150 7MP NIKON COOLPIX S500
KODAK EASYSHARE ONE CASIO EXILIM PRO PANASONIC DMC-FX37A 10.1MP NIKON COOLPIX S200 SONY DSCP150 7MP CANON EOS 1D NIKON COOLPIX 8400 NIKON COOLPIX L1 NIKON COOLPIX P90 SONY ALPHA A700K

Battery

CANON POWERSHOT SD990IS NIKON COOLPIX S4 NIKON COOLPIX S500 NIKON COOLPIX L3 KODAK EASYSHARE C653 HP PHOTOSMART M537 NIKON COOLPIX P90 NIKON COOLPIX P6000 NIKON COOLPIX L4 SONY ALPHA A700K

Compression

CANON POWERSHOT A620 CANON POWERSHOT SD300 KODAK EASYSHARE ZD710 CANON POWERSHOT S100 NIKON COOLPIX 8700 NIKON COOLPIX 4300 NIKON COOLPIX L3 NIKON COOLPIX P50 CANON POWERSHOT S230 CANON POWERSHOT A95

Overall Quality

NIKON COOLPIX L1 PANASONIC DMC-FX37A 10.1MP CANON POWERSHOT A990IS NIKON COOLPIX P90 HP PHOTOSMART M537 CANON POWERSHOT A70 NIKON COOLPIX 8800 FUJIFILM FINEPIX A330 KODAK EASYSHARE C653 OLYMPIS STYLUS 550
a feature should reect the role a feature plays in inuencing a customer. The IF metric measures the agreement between the overall ranking and feature-specic ranking. The values for IF for different product features are shown in Table 6. They indicate that our product feature-based ranking is consistent with overall ranking (which is veried previously by comparing with expert ranking [21]). In addition, this ratio should not be close to 1 because overall quality is different from feature quality. Table 6 also tells us that lens (Digital Camera) and size (Television) are the leading factors inuencing the overall quality of a product. Table 7 shows the top 10 cameras according to the overall ranking and featurespecic ranking. These results are consistent with the arguments made earlier.
a methodology and algorithm to rank products based on their features using customer reviews. First, we manually dene a set of product features that are of interest to the customers. We then identify subjective and comparative sentences in reviews using text mining techniques. Using these, we construct a feature-specic product graph that reects the relative quality of products. By mining this graph using a page-rank like algorithm(pRank), we are able to rank products. We implement our ranking methodology on two popular product categories (Digital Camera and Television) using customer reviews from Amazon.com. We believe our ranking methodology is useful for customers who are interested in specic product features, since it summarizes the opinions and experiences of thousands of customers.

Conclusion

6 Acknowledgement
This work was supported in part by NSF grants CNS0551639, HECURA CCF-0621443, OCI 0956311, SDCI OCI-0724599, OCI-0956311, IIS-0905205, and DOE grant DE-FG02-08ER25848. 8
Recent trends have indicated that large numbers of customers are switching to online shopping. Online customer reviews are an unbiased indicator of the quality of a product. However, it is difcult for users to read all reviews and perform a fair comparison. We describe

References

[1] Forrester research, http://www.comscore. com/Press_Events/Press_Releases/2007/ 11/Online_Consumer_Reviews_Impact_ Offline_Purchasing_Behavior. [2] comscore and Kelsey, http://www. shop.org/c/journal_articles/ view_article_content?groupId= 1&articleId702&version=1.0. [3] M. Hu and B. Liu, Mining and Summarizing Customer Reviews, Proceedings of the 10th ACM International Conference on Knowledge Discovery and Data Mining (SIGKDD-2004), 8 (2004), pp. 168174. [4] M. Hu and B. Liu, Mining Opinion Features in Customer Reviews, Proceedings of the 19th National Conference on Articial Intelligence., 7 (2004), pp. 755-760. [5] A. Popescu and O. Etzioni, Extracting product features and opinions from reviews, Proceedings of the conference on Human Language Technology and Empirical Methods in Natural Language Processing., (2005), pp. 339-346. [6] B. Liu, M. Hu, and J. Cheng, Opinion Observer: Analyzing and Comparing Opinions, WWW., 5 (2005), pp. 342351. [7] S. Kim, P. Pantel, T. Chklovski, and M. Pennacchiotti, Automatically Assessing Review Helpfulness, EMNLP., 7 (2006), pp. 423-430. [8] B. He, C. Macdonald, J. He, and I. Ounis, An Effective Statistical Approach to Blog Post Opinion Retrieval, CIKM., 10 (2008), pp. 1063-1069. [9] B. Liu, Sentiment Analysis and Subjectivity, to appear in Handbook of Natural Language Processing, Second Edition, 2010. [10] B. Pang, L. Lee, and S. Vaithyanathan, Thumbs up? Sentiment Classication using Machine Learning Techniques, EMNLP., (2002), pp. 79-86. [11] J. Reynar and A. Ratnaparkhi, A Maximum Entropy Approach to Identifying Sentence Boundaries., In Proceedings of the Fifth Conference on Applied Natural Language Processing , March 31-April 3, 1997. Washington, D.C. [12] X. Phan, CRFTagger: CRF English POS Tagger, http: //crftagger.sourceforge.net/, 2006. [13] G. A. Mille, C. Fellbaum, etc., WordNet: An Electronic Lexical Database, 5 (1998),MIT Press. [14] M. Hu, and B. Liu, Opinion extraction and summarization on the web, AAAI., 7 (2006), pp. 1621-1624. [15] B. Liu, X. Li, W. S. Lee, and P. S.Yu, Text Classication by Labeling Words, AAAI., 2004, pp. 425-430.

[16] N. Jindal, and B. Liu, Identifying Comparative Sentences in Text Documents, SIGIR., 8 (2006), pp. 244-251. [17] N. Jindal, and B. Liu, Mining Comparative Sentences and Relations, AAAI., (2006), pp. 1331-1336. [18] V. Hatzivassiloglou, and K. R. McKeown, Predicting the semantic orientation of adjectives, Proceedings of the eighth conference on European chapter of the Association for Computational Linguistics, 1997, pp. 174-181. [19] MPQA corpus http://www.cs.pitt.edu/mpqa, 2002. [20] L. Page, S. Brin, R. Motwani, and T. Winograd, The PageRank citation ranking: Bringing order to the Web, Stanford University, 1999. [21] K. Zhang, R. Narayanan, and A. Choudhary, Mining Online Customer Reviews for Ranking Products, Technical Report, EECS department, Northwestern University, 2009. [22] S. Arora, M. Joshi, and C. P. Rose, Identifying Types of Claims in Online Customer Reviews, NAACL HLT., 6 (2009), pp. 37-40. [23] Y. Lu, C. Zhai, and N. Sundaresan, Rated Aspect Summarization of Short Comments, WWW., 4 (2009), pp. 131140.