Patent Grant
9020811
The present disclosure relates generally to searching documents and, more specifically, to searching for information within an audio file.
The statements in this section merely provide background information related to the present disclosure and may not constitute prior art.
Searching text documents using a search engine is well known. Searching takes place using the text within the document to identify relevant search results.
Often times, businesses receive a large quantity of non-text-based material. The non-text material may be referred to as image text documents if the image contains a number of words. Image text documents are images of words but the words are not searchable using a search engine. The text appears as a picture or image only and, thus, cannot be searched. The image text files may originate from various sources including faxes, forms, charts, diagrams, pictures, and the like. Often times, metadata may be stored with the image to help identify the image. The metadata may contain various titles, key people, or the like. The metadata may be manually entered for each document.
Audio files are another way in which a business may receive or store material. Audio files may be generated in various ways from voicemails and audio tracks of videos. Typically, businesses have no way to search audio files or the content therein.
When searching image text documents, only the metadata itself is searched. The search results may have limited accuracy due to the limited nature of the metadata. Thus, the documents found in the search may be limited because there is not enough identifying information in the metadata. Consequently, desirable information may be overlooked. Therefore, it is desirable to improve the results of search engines when faced with image text documents.
The present disclosure is particularly suitable for various types of enterprises that have a large number of documents. The disclosure allows improved search results by converting audio text files to text documents and searching using the text documents.
In one aspect of the disclosure, a method includes processing audio text files generated from audio files comprising linguistically analyzing the audio text files within a context of multiple lexicons to form edited text results, creating reflection files corresponding to the audio text files from the edited text results, storing the reflection files in a reflection repository, processing the reflection files, and displaying a first reflection file from the reflection files or a first audio text file from the audio text files in response to processing.
In a further aspect of the disclosure, a system for searching audio text files includes a content repository storing audio text files. A text transformer linguistically analyzes the audio text files within a context of multiple lexicons to form edited text results and creates a reflection repository having reflection files therein corresponding to the audio text files from the edited text results. A processing tool processes the reflection files and a user device displays a first reflection file from the reflection files or a first audio text file from the audio files in response to processing.
Further areas of applicability will become apparent from the description provided herein. It should be understood that the description and specific examples are intended for purposes of illustration only and are not intended to limit the scope of the present disclosure.
The drawings described herein are for illustration purposes only and are not intended to limit the scope of the present disclosure in any way.
The following description is merely exemplary in nature and is not intended to limit the present disclosure, application, or uses. For purposes of clarity, the same reference numbers will be used in the drawings to identify similar elements. As used herein, the term module refers to an Application Specific Integrated Circuit (ASIC), an electronic circuit, a processor (shared, dedicated, or group) and memory that execute one or more software or firmware programs, a combinational logic circuit, and/or other suitable components that provide the described functionality. As used herein, the phrase at least one of A, B, and C should be construed to mean a logical (A or B or C), using a non-exclusive logical OR. It should be understood that steps within a method may be executed in different order without altering the principles of the present disclosure.
Referring now to
An image text transformer 14 is in communication with the content repository 12. The image text transformer 14 includes a rover 16 for extracting and collecting images from the content repository 12. One suitable example of a rover is Syscom's Meta Image Rover.
The text transformer 14 may include an optical character recognition (OCR) module 18. An optional position for the OCR module 18 is illustrated outside the image text transformer 14. That is, some optical character results may be pre-generated or communicated to the image text transformer 14. From here on, reference numeral 18 will be used to refer to the OCR module regardless of position. Optical character recognition module 18 forms optical character recognition results and creates reflection files in response to the results. A translation module 19 may also be included within the text transformer 14 or outside the text transformer 14 as translation module 19′. The translation module 19, 19′ may process the optical character recognition results and change the optical character recognition results from a first language to a second language. This may take place before further editing or spell-checking. Of course, translation may take place after editing, spell-checking and the like. However, the steps of editing, spell-checking, grammar-checking and the like may take place in a different language and then translated to English, in this example.
The optical character recognition results may be edited or changed in the editing module 20. As will be described below, the editing module 20 may include various types of editing, including linguistically analyzing the optical character recognition results, spell checking the optical character recognition results, and grammar checking the optical character recognition results. Various types of editing or no editing of the optical character recognition results may take place depending on the particular system.
The optical character recognition module may perform OCR on images in the content repository 12. Also, the OCR may be reused from another application such as facsimile recognition. One example is, when facsimile recognition performs OCR, only certain recognized fields are retained to form an index. The remaining OCR results are discarded. The present disclosure may use the OCR results to form or add to the reflection repository 22.
The present disclosure may also be applicable to audio text files that originate from audio files. An audio recognition module 21 and 21′ are illustrated within the text transformer 14 and outside of the image text transformer 14. The audio recognition module 21, 21′ is used to convert an audio file to an audio text file. This may be performed using various languages with translation to the processing language by the translation module 19. An audio text file originates from an audio file such as an analog or digital audio file. Various sources such as digital recorders, tape records, videos and the like may originate audio files. The audio recognition module 21, 21′ converts the audio file to an audio text file using voice recognition.
The text transformer 14 may also include an audio text rover 23. The audio text rover 23 may be used for extracting the audio text files from the content repository 12. The audio text rover 23 provides the audio text files to the text transformer 14 from the content repository 12 for further processing as described below.
The text transformer 14 generates reflection files that are stored in the reflection repository 22. The reflector files may correspond to audio text files or image text documents. The reflection repository 22 may be a separate memory storage device than the content repository 12. The reflection repository 22 may be combined with the content repository 12 in terms of physical location. Also, the reflected documents may also be stored together with their respective document in the content repository. That is, a reflected document and the image text document or audio text file may be linked. The documents and files may be linked so that the same security for the image or audio text files is provided to the reflected documents. The image text documents or audio text files may be stored in the reflection repository 22 with a unique file extension in a form that is compatible for crawling and indexing by various search engines. A suitable search engine is Omnifind's® crawler/indexer or Google®.
The reflection document may be in various formats. For example, the reflection document may be in a rich text format (.rtf), a .pdf format or a .txt format. In a .txt format, the formatting may be removed. This may allow the reflection document to occupy less memory space.
A user device 30, having a display 32, may also be coupled to the text transformer 14. The user device 30 may be a networked computer. The user device 30 may be used to modify batches and control the functioning of the text transformer 14.
The user device 30 may also contain a content viewer 36, a processing tool 38, and search engine 40. The content viewer 36 and the processing tool 38 may be incorporated within the user device 30 as software. The processing tool may include a search engine 40, a text analytics module 42 and/or a search and text analytics module 44. The search engine 40 may be various types of search engines that are used to crawl, parse and index the reflected documents stored in the reflection repository 22. The search engine 40 searches the various reflected documents and returns search results. The text analytics module 42 may be used to analyze the text data. For example, text analytics may compile a quantity of various types of information, including proximity to other words, the number of occurrences within a text and other statistical relationships. The text analytics module 42 may perform linguistic analysis as well as indexing. It should be noted that text analysis typically performs a type of searching. However, the searching mentioned above may be performed in addition to the searching of text analysis. This is set forth in the search and text analytics module 44. The search and text analytics module 44 may provide searching in a crawl-type environment as well as searching in a text analytics environment. Both searching and text analytics may be returned from the processing tool 38.
The content viewer 36 allows the user to generate various views based upon the search results from the processing tool 38. The content viewer 36 may, for example, allow the user 32 to display the reflected document from the reflection repository 22 or the image or audio text files from the content repository 12. Further, the content viewer 36 may allow both the reflected document and the image text document to be viewed side by side. This will be further described in
Multiple lexicons 50 may be used in the editing module 20. The lexicon 50 may be derived from various sources including common words, proper names, client-specific lexicons (custom), application specific lexicons and industry-specific lexicons.
Referring now to
In step 102, the various batches are managed. Batches define a logical unit of work or set of documents upon which the system services will operate. The system will operate on the set of documents in a batch sequentially. Managing the batches also includes coordinating threads to take advantage of processing schedules using multiple CPUs and manage memory. The system may operate upon multiple batches in parallel. The system also allows administrators to coordinate system execution time frames and batch sizes. Managing batches 102 may also include staging the batches including kicking off the batch of work to the staging phase.
In step 104, the batches may be staged. If the device is on a tape device or in off-line storage, the documents may be pre-fetched. By pre-fetching documents, tape-stored documents or off-line storage avoids a subsequent delay time during the processing. Staging may also determine whether or not the document is within a database table. That is, the documents may be within a database instead of being within a file system. If the files are within a database, segments may be built. The rows of data in the database (BLOBs) may be used to create a file if this document is within a database. The staging phase 104 may also track the results of the staging phase, including the quantity of documents, time to execute, size of documents, and the number of errors.
In step 106, roving of the content repository using the text transformer may be managed. Roving is used to translate an image to text. Various parameters may be set for performing the roving. Managing the roving may coordinate threads to take advantage of processing schedules, use of multiple CPUs, and manage memory. The system may operate upon multiple batches in parallel. The system may allow administrators to define the level of parallel roving to manage. Managing the roving may also include splitting the work up into multiple streams to be operated upon.
In box 108, several steps that illustrate roving are set forth. The roving may take place using the image rover 16. In step 110, the image files are retrieved by the image text transformer 14 from the content repository 12. The metadata may also be retrieved in step 112. In step 114, the recognition of a particular form may be performed. For example, if the document is a particular type of document, this type of form may be recognized in the form-recognition step of 114. In step 116, if the document includes bar codes or supports bar-code recognition, a bar code transformation process may take place. The bar codes may, thus, be interpreted also in step 116. In step 118, an optical character recognition (OCR) process may be performed on the documents. The OCR process may use a number of types of processing.
As mentioned above, the OCR may be performed as a result of another process such as a facsimile recognition process that typically discards OCR results after field indexing is performed.
Steps 120-136 perform editing steps. In step 120, the OCR results may be translated from a first language to a second language. In step 128, lexicons may be formed. The lexicons may be derived from extractions of a customer's structure data or may be a third-party linguistic analytical engine. Common lexicons may also be used as well as standard proper lexicons. Thus, various lexicons may be used to improve the optical character recognition results and/or the translated results. Of course, translation may occur after text resolution. In step 130, the lexicons are used to perform text resolution from the results of the optical character recognition. The text resolution is part of an editing block 126.
In step 132, linguistic analysis from a third-party analytic engine may also be used to help resolve the results of the optical character recognition. Further editing may include a grammar check 134. Various types of grammar checking are known, such as those used in word processors. In step 136, further editing may also take place. The spell check may be similar to that set forth in a word processor. It should be noted that the editing in steps 130-136 may be split apart and processed in parallel batches. An edit manager may be used to schedule and perform the edits illustrated in steps 130-136. After the editing steps in box 126, the text transformer in step 128 may encode the edited image text documents for storage in the reflection repository 22 as reflection documents. In step 130, the reflection documents are stored in the reflection repository. A reflection repository may also store metadata.
The linguistic analysis, grammar check and spell check may be used to change the reflected document or may also be used to change the metadata. That is, if a grammar error in linguistic or spelling check is noted, the reflection document itself may not be modified but the metadata may be modified instead. Another alternative is for the end of the document to be modified with the correct grammar, spelling check, or linguistic change within the document but at the end of the document at the point of the change within the document.
The encoding and storing may be formed so that the reflected documents may easily be searched. The storing, as mentioned above, may also take place in the content repository. The reflected documents may be linked together with the image text document. Linking may be performed so that the same security for the image text document is applied to the reflected document. Each of the documents may have different levels of security, depending on the information contained therein.
Referring now to
In step 138, roving of the content repository using the text transformer may be managed. Roving involves retrieving the audio files and performing audio recognition thereon. The management of roving may include splitting the work into multiple streams to be operated upon by the text transformer.
Box 140 illustrates the steps to be performed in the roving process. Step 142 retrieves the audio file from the content repository. Step 144 retrieves any metadata associated with the audio files. This may include the author or speaker in the audio file, the time of the audio file, and other indications of origin of the audio file. The metadata may also include the subject and key words associated with the audio file. In step 146, audio recognition of the audio file is performed. Audio recognition may, for example, be voice recognition. As is illustrated in
After the initial audio recognition, the audio text files may be edited and translated in step 148. As mentioned above, translation may take place after grammar and spell-checking. A series of steps performed during editing in step 150 is set forth. As described above, lexicons may be formed and used in the processing. Lexicons may be derived from extractions of customer's structured data or may be a third-party linguistic and analytical tool. Within the context of multiple lexicons, the text document may be improved in step 154. In step 156, linguistic analysis may also be performed.
In step 156, linguistic analysis may also be performed on the audio text file to form edited results. For example, a third-party analytical engine may be used as described above in step 132 to form further edited results. Further, steps 158 for grammar-checking and 160 for spell-checking correspond to steps 134 and 136, respectively. As mentioned above, various types of grammar and spell-checking similar to that set forth in a typical word processor may be used.
In steps 162 and 164, the edited results may be encoded and stored in the reflection repository. These steps are similar to the steps set forth above relative to the image text files. Step 162 encodes the edited results for storage within the reflection repository to form the reflection documents. Step 164 stores the reflection documents in the reflection repository. Metadata may also be stored with the reflection documents. The reflection documents correspond to the audio text files.
Referring now to
In step 186, the user initiates the processing tool of
In step 196, the reflected documents may be viewed. In step 198, the original image text document may also be viewed. It should be noted that the reflected document may be viewed together with the original document as illustrated in
In step 200, the documents may be manipulated. That is, various portions of the document may be scrolled, zoomed, panned, rotated, fitted, sized, or negatives made. Also, further manipulations may be performed by annotating, copying, sending, printing, or changing the properties of the document. Spell checking and grammar checking may also be performed after viewing the reflected document. The changes may be performed in the reflected document or in the metadata. Other manipulations include synchronizing the document, highlighting, dual highlighting, scrolling, dual scrolling, viewing edits, and the like. The viewing edits may be performed by looking at both documents and editing the text of the reflected document.
Referring now to
Referring now to
Referring now to
Referring now to
Those skilled in the art can now appreciate from the foregoing description that the broad teachings of the disclosure can be implemented in a variety of forms. Therefore, while this disclosure includes particular examples, the true scope of the disclosure should not be so limited since other modifications will become apparent to the skilled practitioner upon a study of the drawings, the specification and the following claims.
This application is a continuation-in-part of U.S. patent application Ser. No. 11/871,349, filed Oct. 12, 2007, which claims priority to U.S. Provisional Application No. 60/851,799, filed on Oct. 13, 2006. The entire disclosures of each of the above applications are incorporated herein by reference.
| Number | Name | Date | Kind |
|---|---|---|---|
| 5923792 | Shyu et al. | Jul 1999 | A |
| 5933531 | Lorie | Aug 1999 | A |
| 6047251 | Pon et al. | Apr 2000 | A |
| 6125362 | Elworthy | Sep 2000 | A |
| 6175663 | Huang | Jan 2001 | B1 |
| 6577755 | Lorie | Jun 2003 | B1 |
| 6608930 | Agnihotri et al. | Aug 2003 | B1 |
| 6766069 | Dance et al. | Jul 2004 | B1 |
| 6937766 | Wilf et al. | Aug 2005 | B1 |
| 6964022 | Snowdon et al. | Nov 2005 | B2 |
| 7287024 | Snowdon et al. | Oct 2007 | B2 |
| 7320128 | Hirai | Jan 2008 | B2 |
| 7469063 | Koyama et al. | Dec 2008 | B2 |
| 7734636 | Handley | Jun 2010 | B2 |
| 7769249 | Kasatani | Aug 2010 | B2 |
| 7818215 | King et al. | Oct 2010 | B2 |
| 8214387 | King et al. | Jul 2012 | B2 |
| 8249309 | Kurzweil et al. | Aug 2012 | B2 |
| 8290270 | Bayer et al. | Oct 2012 | B2 |
| 8339619 | Misawa | Dec 2012 | B2 |
| 20020103834 | Thompson et al. | Aug 2002 | A1 |
| 20020120647 | Amano | Aug 2002 | A1 |
| 20020126120 | Snowdon et al. | Sep 2002 | A1 |
| 20040213458 | Kanatsu | Oct 2004 | A1 |
| 20040220962 | Kaneda | Nov 2004 | A1 |
| 20060053097 | King et al. | Mar 2006 | A1 |
| 20060210162 | Sato | Sep 2006 | A1 |
| 20070072578 | Lee et al. | Mar 2007 | A1 |
| 20070168382 | Tillberg et al. | Jul 2007 | A1 |
| 20080002893 | Vincent et al. | Jan 2008 | A1 |
| 20080002914 | Vincent et al. | Jan 2008 | A1 |
| 20080002916 | Vincent et al. | Jan 2008 | A1 |
| Entry |
|---|
| Chellapilla, K., et al., “Combining Multiple Classifiers for Faster Optical Character Recognition,” Microsoft Research, Feb. 2006, 10 Pages. |
| Levin et al., “Dynamic Planar Warping for Optical Character Recognition,” IEEE International Conference on Acoustics, Speech, and Signal Processing, vol. 3, Mar. 23-26, 1992, pp. III-149-III-152. |
| Saiga et al., “An OCR System for Business Cards,” IEEE Comput. Soc., pp. 802-805, 1993. |
| Luo et al., “Design and implementation of a card reader based on build-in camera,” IEEE, vol. 1, pp. 417-420, 2004. |
| Sibun “Language Determination: Natural Language Processing From Scanned Document Images”; 4th Conf. on Applied Natural Language Processing. Stuttgart, Oct. 1994; pp. 15-21. |
| Likforman-Sulem “Proper names extraction from fax images combining textual and image features”, Document Analysis and Recognition, 2003. Proceedings. Seventh International Conference on DOI: 10.1109/ICDAR.2003.1227724, Publication Year: 2003 , pp. 545-549 vol. 1. |
| P. Sibun and A. L. Spitz, “Language Determination: Natural Language Processing from Scanned Document Images.” Proceedings of the 4thConference on Applied Natural Language Processing, Stuttgart Germany , pp. 15-21, 1994. |
| (Tong & Evans 1996) Tong, X and Evans, D. A., A Statistical Approach to Automatic OCR Error Correction in Context. In: Proceedings of the fourth Workshop on Very LargeCorpora, 88-100, Copenhagen, Denmark, 1996. |
| Sato T. et al: “Video OCR for digital news archive” Content-Based Access of Image and Video Database, 1998. Proceedings., 1998 IEEE International Workshop on Bombay, India Jan. 3, 1998, Los Alamitos, CA USA, IEEE Comput. Soc, US, 1998, pp. 52-60,XP010261519. |
| Hadhoud, M. M. et al: “New trends high resolution image processing” Photonics and its Application, 2004. The fourth workshop on Giza, Egypt May 4, 2004, Piscataway, NJ, USA, IEEE, 2004, pp. 1-23, XP010737924. |
| KMWorld Names WebQL 3.0 in “Trend-Setting Products of 2006,” www.ql2.com/news/press—releases—kmwtsp06—082506.php. |
| QL2 Software Unveils WebQL 3.0, Computing News Press Service © 2006 Computing News :: home page. |
| Business Intelligence Lowdown, “Image Text Search Tool,” Mar. 7, 2006, www.businessintelligencelowdown.com/2006/03/image—text—sear.html. |
| Number | Date | Country | |
|---|---|---|---|
| 20130041892 A1 | Feb 2013 | US |
| Number | Date | Country | |
|---|---|---|---|
| 60851799 | Oct 2006 | US |
| Number | Date | Country | |
|---|---|---|---|
| Parent | 11871349 | Oct 2007 | US |
| Child | 13651541 | US |
