Скачать презентацию Question-Answering of Large News Video Archives CHUA Tat-Seng Скачать презентацию Question-Answering of Large News Video Archives CHUA Tat-Seng

c4de0bc78d6f74fb5ab9c3f90220c4fc.ppt

  • Количество слайдов: 36

Question-Answering of Large News Video Archives CHUA, Tat-Seng, Yang, Hui, Chaisorn, Lekha & Zhao, Question-Answering of Large News Video Archives CHUA, Tat-Seng, Yang, Hui, Chaisorn, Lekha & Zhao, Yun-Long School of Computing National University of Singapore Email: [email protected] nus. edu. sg Web: http: //www. comp. nus. edu. sg/~chuats

Outline of Talk • Introduction and Motivation • News Video Processing & Story Segmentation Outline of Talk • Introduction and Motivation • News Video Processing & Story Segmentation • Video Transcript Correction • Question-answering on News Video • Results • Conclusion

Personalized News Video Retrieval • Infotainment, including news video, is one of the major Personalized News Video Retrieval • Infotainment, including news video, is one of the major applications of MM Technology • In a personalized news video scenario, users “interact” with the system to enquire info such as: o show me latest news video on Iraq “Iraq” o highlight of last nights European football “European football” o Results are time-specific • Users increasingly want to see video news, supplemented with audio and text o and summarized to as much detail as is necessary • In a more futuristic setup, these will be accomplished through “natural” human-oriented I/O 3

Issues to Resolve • Imprecision of users queries o “highlight of football match last Issues to Resolve • Imprecision of users queries o “highlight of football match last night? ” • Extraction of semantic contents of video: o Multi-modality o Multi-sources • Segmentation of news video into story units with genre classifications • Summarization of info for viewing at different level of details 4

What Kinds of Data Do we Have? • Most research in the past has What Kinds of Data Do we Have? • Most research in the past has looked into only one source o Example, video and its accompanying audio track, + ASR • In most real-life applications, information is readily available in multiple sources: o o Broadcast news -- video and audio Web-based news articles (by news stations) On-line wired news (by news agencies) Other general resources: ontologies, dictionary etc… • Other types of info increasingly used in IR community: o User models: query logs, user profiles etc. • A challenge in developing usable systems. . How to use these available data effectively In co-training/ testing type framework? ? Ignoring these obvious data resources will result in unsatisfactory 5 solutions.

Outline of Our Approach • In this talk, I will describe our approach in Outline of Our Approach • In this talk, I will describe our approach in developing systems to handle large scale video corpuses – TREC video • Sources of data used: o News video itself: visual, audio features, ASR o External sources: on-line news articles of the same period o General resources – ontology of countries, dictionary - WORDNET • Approach (see architecture): 6

Overview of QA on News Video Stage 1: Stage 2: Stage 3: Stage 4: Overview of QA on News Video Stage 1: Stage 2: Stage 3: Stage 4: Stage 5: System Architecture of Video. QA Stage 6 7

Outline of Talk • Introduction and Motivation • News Video Processing & Story Segmentation Outline of Talk • Introduction and Motivation • News Video Processing & Story Segmentation • Video Transcript Correction • Question-answering on News Video • Results • Conclusion

Video Story Segmentation for News Video • • Intro First basic problem: break the Video Story Segmentation for News Video • • Intro First basic problem: break the news video into meaningful units based on stories. Issues: o How to classify shots into the correct class/category? o How to detect story boundaries? Most news adopt the structure similar to CNN’s (? ) News Com 1 News Finance News Com 2 Sports News Weather 9

Video Story Segmentation for News Video -2 • To help alleviate the estimation problem Video Story Segmentation for News Video -2 • To help alleviate the estimation problem in statistical learning, we adopt a two stage process: o Stage 1: Shot classification o Stage 2: Scene segmentation & classification • The set of features considered o Visual (color histogram, b/g change) o Temporal [Motion activity, Audio type, Shot duration, speaker change] o Mid-Level [# of Faces, Shot type, # of Text Lines, and textposition, cue phrases] 10

Stage 1: Shot Classification • • Divideo sequence into shots Consider 13 categories of Stage 1: Shot Classification • • Divideo sequence into shots Consider 13 categories of shots • Perform classification using Decision Tree (SEE 6. 0) o o o Intro/Highlight Anchor; 2 -Anchor; Meeting; Speech Still image shot; Text Scene Sports; Live reporting Finance; Weather; Commercial; Special 11

Stage 2: Scene Detection • Employ Hidden Markov Model (HMM) to detect story • Stage 2: Scene Detection • Employ Hidden Markov Model (HMM) to detect story • boundaries Features (sequence level features) used at this stage: o o Shot classes – shot tags Scene change [c/u] Speaker change [c/u] Cue phrases at the beginning of new stories • Input to HMM: [1 cc 1 uu 1 cu. . 2 cc 4 c 4 uu 6 uu …. 2 cc …. ] • Tested on 120 hours of TREC video and achieve • around 76% in F 1 accuracy in story segmentation TREC data may be down-loaded from TREC web sites later (? ) (Chaisorn & Chua et al, ICME’ 02, WWW Journal’ 02, TREC’ 03) 12

Outline of Talk • Introduction and motivation • News Video Processing & Story Segmentation Outline of Talk • Introduction and motivation • News Video Processing & Story Segmentation • Video Transcript Correction • Question-answering on News Video • Results • Conclusion

Text Transcript: from Speech to Text • Need accurate transcript for QA • Performance Text Transcript: from Speech to Text • Need accurate transcript for QA • Performance of speech recognition system • How to correct errors in ATs? • o not a problem for document or story retrieval o o Accuracy about 80% for news Most errors are named entities – likely answer targets (ATs) Most such errors are type substitution homonym problem Examples: pneumonia new area; Tony Blair Teddy Bear use phonetic sound matching to correct the errors o May use confusion matrix successfully used in spoken docm retrieval o Problem: low precision match to many irrelevant phrases One solution: limit scope of phonetic sound match o By utilizing on-line text news of same period (extract base noun 14 phrases and named entities) – reasonable

Use of External Resource to correct Speech Errors • • • Extract all ATs Use of External Resource to correct Speech Errors • • • Extract all ATs from on-line news articles, Ai = (ai 1, . . aiq) Given video transcript Ti with a list of terms (ti 1, . . , tip) The basic problem is then to select an aik Ai to replace a sequence of terms sj Ti that maximizes the probability: where sj contains one or more consecutive terms in Ti • Basic idea: use co-occurrence probabilities & phonetic matching to find most likely aik Ai to replace sequence of terms sj Ti, : a) Extract list of probable ATs using co-occurrence probabilities a) Matching at phonetic syllable level; b) Matching at confusion syllable string level (see Wang & Chua, ACL’ 03) 15

Outline of Talk • Introduction and Motivation • News Video Processing & Story Segmentation Outline of Talk • Introduction and Motivation • News Video Processing & Story Segmentation • Video Transcript Correction • Question-answering on News Video • Results • Conclusion

Overview of QA on News Video System Architecture of Video. QA (Similar to our Overview of QA on News Video System Architecture of Video. QA (Similar to our text-based QA work – Yang & Chua, SIGIR’ 03) 17

Question Processing • Users typical issue short queries (several keywords): o “development in North Question Processing • Users typical issue short queries (several keywords): o “development in North Korea” o “match last night” o Query is ambiguous!! • Example: Analyze the query to extract: o o o Key terms in query Likely answer target NP & NE in query Type of video genre Temporal constraint Duration constraint “football match last night? ” “football”, “match” “football team” (ORG-NAME) “football match” SPORTS LAST-NIGHT 30 seconds (default) 18

Query Reinforcement • The query, however, is ambiguous! o Use on-line news articles to Query Reinforcement • The query, however, is ambiguous! o Use on-line news articles to provide the context (user independent) • Basic Idea: Given original query q(o): o o • Use web (or news sites) and dictionary – Word. Net Find terms (from web articles) co-occur frequently with q(o) Extract semantically related terms from Word. Net Add high probability terms into q(0) to get q(1) Expect q(1) to contain more context terms than q(0) o For the football example: we expect q(1) to also contain terms like: “arsenal”, “inter milan”, “soccer”, etc (the big match last night) 19

Query Reinforcement Another Example • • q(0) = “What are the symptoms of atypical Query Reinforcement Another Example • • q(0) = “What are the symptoms of atypical pneumonia? ” q(1) = “symptoms, pneumonia, virus, spread, fever, cough, breath, doctor” Use q(1) to retrieve a list of news transcripts at story level 20

Candidate Sentence Extraction • For the retrieved transcript Ti, we select sentences Sentij that Candidate Sentence Extraction • For the retrieved transcript Ti, we select sentences Sentij that best match the user query as follows: o o o • • noun phrases, wnj named entities, whj original query words q(0), wcj expanded query words q(1 -0) = q(1) - q(0), wej video genre, wvj Final score is: where αk=1 and wkj = {wnj, whj, wcj, wej, waj, wvj} The top K sentences are selected as the candidate answer sentences based on Sij 21

Outline of Talk • Introduction • News Video Processing & Story Segmentation • Video Outline of Talk • Introduction • News Video Processing & Story Segmentation • Video Transcript Correction • Question-answering on News Video • Results • Conclusion

Results • Use 7 days of CNN news video from 13 -19 Mar 2003 Results • Use 7 days of CNN news video from 13 -19 Mar 2003 o contained a total of 350 minutes of news video o retrieved about 600 news articles per day from the Alta Vista news web site during these 7 days • Designed 40 factoid questions o 28 general questions that are asked everyday o 12 questions are date-specific o Give a total of 208 questions • Results Transcript Correct Answers Accuracy without error correction 116 55. 8% with error correction 153 73. 6% (To present in ACM Multimedia ’ 03) 23

Results -- Example • • Query: “What are the symptoms of atypical pneumonia? ”, Results -- Example • • Query: “What are the symptoms of atypical pneumonia? ”, the 3 -sentence window selected by the QA engine is o S 1: He and his two companions are now in isolation and the one hundred and fifty five passengers on the flight were briefly quarantined. o S 2: Symptoms include high fever, coughing, shortness of breath and difficulty breathing. o S 3: But health officials say there's no reason to panic. • The video summary example (4 shots) is: 24

Outline of Talk • Introduction • News Video Processing & Story Segmentation • Video Outline of Talk • Introduction • News Video Processing & Story Segmentation • Video Transcript Correction • Question-answering on News Video • Results • Conclusion

Related Work • Research in correcting speech recognition errors (ACL’ 03, EMNLP’ 02) • Related Work • Research in correcting speech recognition errors (ACL’ 03, EMNLP’ 02) • News story and dialogue segmentation (Columbia U) (ICME’ 03, ACL’ 03) • Question-answering in text (TREC’ 02, SIGIR’ 03) • • • Infomedia Project o Uses multi-modality features effectively, esp speech o Insufficient emphasis on external resources Works on Video-TREC - Large scale testing Collaboration with Ramesh jain (Georgia Tech) as part of Video Tagging Project o Employ TV-Anytime metadata for news (collaborate with ETRI Korea) o Automatic tagging of TV-Anytime metadata, and use it as basis 26 for video QA

Summary • Works are preliminary o Many processes needs to be automated • Participating Summary • Works are preliminary o Many processes needs to be automated • Participating in this year’s Video-TREC and test on large scale corpuses (120 hours of news video) o On both story segmentation and retrieval • Experience: o Story Segmentation: content features are important, text or ASR feature less important o Retrieval: Text or ASR is important; content features help in enhancing precision • Current Work: o Build appropriate meta model to encode domain knowledge o Use higher order statistics to analyze data • KEY MESSAGE– Must incorporate domain model and utilize multi-modality, multi-source information 27

THANK YOU 28 THANK YOU 28

Question classification and possible video genres Answer Target Likely Video Genre Example Human Anchor, Question classification and possible video genres Answer Target Likely Video Genre Example Human Anchor, meeting, speech, General-news Who is the Secretary of State of the United States? Location Live report, Anchor, General-news Where is Saddam Hussein hiding? Organization Live report, anchor Which hospital is the center for SARS treatment in Singapore? Time Anchor, General-news When did the Iraq war start? Number Finance What is the expected GDP of Singapore this year? Sports, Text-scene How many points did Yao Ming score? Weather, Text-scene What is the highest temperature tomorrow? Object Anchor, Still-image, Textscene Which kinds of bombs are used in the current Iraq war? Description Anchor, Text-scene What does SARS stand for? 29

Question analysis Question q(0) What is the score of the football match last night? Question analysis Question q(0) What is the score of the football match last night? What are the symptoms of atypical pneumonia? score, football, match, last, symptoms atypical night pneumonia n football match, last night symptom, atypical pneumonia h football atypical pneumonia Answer Target Number Description Video Genre Sports, Text-scene General News 30

List of Questions 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. List of Questions 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. Who is the British Prime Minister? Who is elected to be China's President? Who is the President of the United States? What is the name of the former Premier of China? What is the name of the new Premier of China? Who will pay the heaviest tallies? Who was arrested in Pakistan? Which musician called off his US tour? When will NASA resume shuttle flights? When will Germany, France and Russia meet? When is the funeral of Djin. Djic? Which are three countries involved in the summit today? Where was the summit held? Which city is the capital of Central African Republic? Which are three major war opponent countries? To whom US withdrew the aid offer? Which country vowed to veto the resolution today? Which country's compromise proposal was rejected by US? Where is Kashmir Hotel? Where did Iraq invite the chief weapons inspectors to? 31

List of Questions – cont. 21. 22. 23. 24. 25. 26. 27. 28. 29. List of Questions – cont. 21. 22. 23. 24. 25. 26. 27. 28. 29. 30. 31. 32. 33. 34. 35. 36. 37. 38. 39. 40. Which city has the largest anti war demonstration? Where did a AL QUEDA suspect arrested? How many people attended the rally in San Francisco? What is the cost of war? How many people were killed in a Kashmir Hotel? How many people participated in the rally in Madrid? How many people were killed by the new pneumonia? What are the symptoms of the atypical pneumonia? What sanction did President Bush lift? What was the name of the space shuttle broken apart in February? Which rally shows the support for President Bush? What is the official name for the mysterious pneumonia? Which company tests their new passenger profiling system? Name one Jewish holiday. What is British stance? How did Serbs Prime Minister die? How is the anti-war protest in Madrid? How is tomorrow's weather? What is the conflict between US and Turkey? What does the WHO call the new pneumonia? 32

Some Remarks on Story Segmentation Task • Our 2 -stage approach helps alleviate the Some Remarks on Story Segmentation Task • Our 2 -stage approach helps alleviate the statistical estimation problem – requires less training data • Similar works done in Columbia U o Using maximum entropy method o For video segmentation (ICME’ 03) and dialogue segmentation (ACL’ 03) o Achieves similar performance • Our current work: o Integration of multiple machine learning methods: HMM, ME, heuristic rule methods, and co-training approach o Fusion of multiple modal features: visual/audio features, text (speech to text), meta-data + domain knowledge o Note: Use only text feature (ASR) performs badly 33

Multi-tier mapping (Wang, Chua, ACL’ 03) • We perform matching at 2 levels to Multi-tier mapping (Wang, Chua, ACL’ 03) • We perform matching at 2 levels to find the most likely aik Ai to replace the sequence of terms sj Ti, : a) Phonetic syllable level; b) confusion syllable string level • Recall Precision At each level, we compute: o LCS(qi, cj): gives longest common subsequence (LCS) match between aik and sj at phonetic syllable level in the order of their occurrence o Mk == I for Levels a and b match; and == coefficients of confusion matrix at Level c match 34

Query Reinforcement • The query, however, is ambiguous! o Use on-line news articles to Query Reinforcement • The query, however, is ambiguous! o Use on-line news articles to provide the context (user independent) • Basic Idea: Given original query q(o): o o o • Go to web (or news sites) to retrieve top N documents Extract terms with high co-location probabilities with q(o), Cq Extract semantically related terms from Word. Net, Gq & Sq Extra terms to be added: Kq = Cq + (Gq Sq) (q(1)= q(0)+{top m terms Kq with weights>=σ} Expect q(1) to contain more context terms than q(0) o For the football example: expect q(1) to also contain terms like: “real madrid”, “manchester united”, “soccer” 35

Query Reinforcement Another example • • q(0) = “What are the symptoms of atypical Query Reinforcement Another example • • q(0) = “What are the symptoms of atypical pneumonia? ” q(1) = “symptoms, pneumonia, virus, spread, fever, cough, breath, doctor” Use q(1) to retrieve a list of news transcripts at story level 36