Towards Constructing a Chinese Information Extraction System to Support Innovations in Library Services World Library and Information Congress: 72nd IFLA General Conference and Council, August 2006, Seoul, Korea Library of Chinese Academy of Sciences Zhang Zhixiong, Li Sa, Wu Zhengxin, Lin Ying
outline 1. Introduction 2. What is IE (Information Extraction)? 3. Potential functions in Innovations of Library Services 4. Constructing a Chinese Information Extraction System 5. Tests and Evaluation
1. Introduction Library of Library of Chinese Academy of Sciences –Now changing the name to National Science Library of China –about 400 staffs, HQ in Beijing, 3 branches in Lanzhou, Chengdu, Wuhan, –serve 90 CAS research institutes across the country –in 2001,initiated Chinese National Science Digital Library (CSDL) program
1. Introduction CSDL (Chinese National Science Digital Library ) CSDL (Chinese National Science Digital Library ) –provided abundant digital information resources for users. (e-journals,6000 west,11000 Chinese, in one day) –developed information systems to support networked services.
Union Catalogs & Document Delivery
Federated database search
Digital reference
remote authentication
1. Introduction CSDL (Chinese National Science Digital Library ) CSDL (Chinese National Science Digital Library ) –provided abundant digital information resources for users. (e-journals,6000 west,11000 Chinese, in one day) –developed information systems to support networked services. –Carried out lots of training and propaganda program
1. Introduction CSDL become one of the key research facility to researcher and graduated students of CAS. CSDL become one of the key research facility to researcher and graduated students of CAS. While While –Information requirement of researcher and graduated students changed rapidly –Traditional information retrieval methods is not sufficient
1. Introduction The User of CSDL want to: The User of CSDL want to: –get rid of the information noise –effectively get a comprehensive view of recent development of domain –disclose significant relationships between information The Librarian of CSDL want to: The Librarian of CSDL want to: –improve the service standard of CSDL –turn the digital library into a knowledge repository
1. Introduction Information Extraction (IE) is the emerging technology serves to our needs Information Extraction (IE) is the emerging technology serves to our needs
outline 1. Introduction 2. What is IE (Information Extraction)? 3. Potential functions in Innovations of Library Services 4. Constructing a Chinese Information Extraction System 5. Tests and Evaluation
2.What is IE (Information Extraction)? NLP Research Group, Univsity of Sheffield NLP Research Group, Univsity of Sheffield –Information extraction (IE) is a term that has come to be applied to the activity of automatically extracting pre-specified sorts of information from natural language texts
2.What is IE (Information Extraction)? Dr. Hamish Cunningham Dr. Hamish Cunningham –IE is a process that takes texts (and sometimes speech) as input and produces fixed-format, unambiguous data as output –Input unstructured unstructured free text free text –Output fixed-format fixed-format unambiguous unambiguous
2.What is IE (Information Extraction)? Output (structured information source) can be used for: Output (structured information source) can be used for: –searching –analysis –generating summary –constructing indices
##### ####### NHS TRUST - PATIENT CASE NOTE ########:######### ####### DOB: 1944 CLEF-RMH-Entry-Key: 52A4F6DB2B46E AB 1992 Seen in General Surgical This lady who has had a mastectomy and left open capsulotomy and removal of her prosthesis was seen by me in the clinic today on behalf of XXXXXXXXXXX. She has extensive bony lymphoedema in her left arm which does not seem to be getting any better although she is more or less reconciled to the problem. The original problem was that she complained of shooting pain in the direction of ulna nerve and although there does not seem to be any evidence of local, regional or distant recurrence the pain itself warrants management in a pain clinic. XXXXXXXXX could be seen in the pain clinic at the XXXXXXX but as this would involve a lot of travelling would like to be treated nearer her home. I wonder whether it would be possible for you to investigate if there is a pain clinic available at XXXXXXXXXXX as I am sure XXXXX could be treated and benefit from its management. I have otherwise arranged for her to be seen in the clinic again in a year's time. There are no signs of recurrence at this time. 5213A4F612F1 IE, A example recurrence no signs of recurrence bony lymphoedema shooting pain in the direction of ulna nerve pain Interventions Problems Problem Site Locations left arm local, regional or distant a year’s time today at this time Time pain clinic clinic pain clinic General Surgical pain clinic mastectomy left open capsulotomy removal of her prosthesis management
IE, A example Extracted Information could be collected… Interventions Problems Problem Site Locations Time recurrence no signs of recurrence bony lymphoedema shooting pain in the direction of ulna nerve pain left arm local, regional or distant a year’s time today at this time pain clinic clinic pain clinic General Surgical pain clinic mastectomy left open capsulotomy removal of her prosthesis management recurrence no signs of recurrence bony lymphoedema shooting pain in the direction of ulna nerve pain left arm local, regional or distant a year’s time today at this time pain clinic clinic pain clinic General Surgical pain clinic mastectomy left open capsulotomy removal of her prosthesis management recurrence no signs of recurrence bony lymphoedema shooting pain in the direction of ulna nerve pain left arm local, regional or distant a year’s time today at this time pain clinic clinic pain clinic General Surgical pain clinic mastectomy left open capsulotomy removal of her prosthesis management
2.What is IE (Information Extraction)? 5 kinds of Information Extraction tasks 5 kinds of Information Extraction tasks –Named Entity recognition (NE) –Coreference resolution (CO) –Template Element construction (TE) –Template Relation construction (TR) –Scenario Template production (ST)
2.What is IE (Information Extraction)? NE is about finding entities NE is about finding entities CO about which entities and references (such as pronouns) refer to the same thing CO about which entities and references (such as pronouns) refer to the same thing TE about what attributes entities have TE about what attributes entities have TR about what relationships between entities there are TR about what relationships between entities there are ST about events that the entities participate in. ST about events that the entities participate in.
2.What is IE (Information Extraction)? Information Extraction will: Information Extraction will: –play a very important role in coping with the huge collections of digital information –bring innovations in library services
outline 1. Introduction 2. What is IE (Information Extraction)? 3. Potential functions in Innovations of Library Services 4. Constructing a Chinese Information Extraction System 5. Tests and Evaluation
3. Potential functions in Innovations of Library Services 1. Automatic annotation and metadata creation –automatic annotation of digital materials –automatic acquisition of metadata –For example, MnM, S-CREAM, AERODAML, SemTag, KIM, hTechsight –ontology-based IE techniques
3. Potential functions in Innovations of Library Services 2. Improving data mining in information analysis –Large-scale data analysis –Detection of many types of evidence –Get enough structured data for analysis
3. Potential functions in Innovations of Library Services 3. Developing knowledge base from free text –statistical and numeric databases –terminological database –fact sheets –SOBA (SmartWeb Ontology-Based Annotation)
3. Potential functions in Innovations of Library Services 4. Generating answers in digital reference system –Most research libraries establish digital reference service –Can we get answers directly from information systems –Natural language QA (Question Answering)
SO … IE is very important IE is very important How to build an IE system (Chinese) How to build an IE system (Chinese) CSDL try to find an effective way CSDL try to find an effective way
outline 1. Introduction 2. What is IE (Information Extraction)? 3. Potential functions in Innovations of Library Services 4. Constructing a Chinese Information Extraction System 5. Tests and Evaluation
4. Constructing a Chinese Information Extraction System A Chinese IE solution A Chinese IE solution –which makes full use of GATE –trying to develop a Chinese IE plug-in to process Chinese information resource based on GATE framework.
4. Constructing a Chinese Information Extraction System GATE GATE – (General Architecture for Text Engineering) –Open Source, Developed from 1995 –GATE, a framework Language Resources (LRs) Language Resources (LRs) Processing Resources (PRs) Processing Resources (PRs) Visual Resources (VRs) Visual Resources (VRs) –ANNIE (A Nearly-New IE system) tokeniser, sentence splitter, POS tagger, gazetteer, finite state transducer and orthomatcher tokeniser, sentence splitter, POS tagger, gazetteer, finite state transducer and orthomatcher
ANNIE Pipeline
GATE: good for English
GATE: Not so good for Chinese
4.Constructing a Chinese Information Extraction System Key difficulties for Chinese information extraction Key difficulties for Chinese information extraction –Chinese tokenizing –Chinese gazetteers –Chinese named entity recognition
Chinese tokenizing English language English language –words are separated by white space and punctuation Chinese Language Chinese Language –without any separation between words
a simple sentence (I am a Chinese) can be broken into several forms with segmenter (I am a Chinese) (I am China person) (I am center country person)
Chinese gazetteers GATE gazetteer lists for English GATE gazetteer lists for English – very abundant GATE gazetteer lists for Chinese process GATE gazetteer lists for Chinese process –simple and short gazetteers such as date, time, organization, location, money, province etc –for a flexible language like Chinese, the list is very limited
Chinese named entity recognition GATE system uses JAPE (a Java Annotation Patterns Engine) rules to recognize NE GATE system uses JAPE (a Java Annotation Patterns Engine) rules to recognize NE
JAPE rules grammar of Chinese is quite different from that of English grammar of Chinese is quite different from that of English the JAPE rules provided by GATE are not suitable for Chinese texts the JAPE rules provided by GATE are not suitable for Chinese texts We need to rewrite JAPE rules to implement Chinese information extraction We need to rewrite JAPE rules to implement Chinese information extraction
Solutions to the problems
three main tasks we have done 1. Integrating ICTCLAS to perform words segmentation
three main tasks we have done 2. Developing Chinese gazetteers to enrich GATE language resources
three main tasks we have done 3. Rewriting JAPE rules to recognize Chinese NE
Chinese JAPE rule Chinese JAPE rule
outline 1. Introduction 2. What is IE (Information Extraction)? 3. Potential functions in Innovations of Library Services 4. Constructing a Chinese Information Extraction System 5. Tests and Evaluation
one years of working, we implemented the system one years of working, we implemented the system carry out experiments carry out experiments
Same piece of article
Our output
Conclusions bring forth a solution for Chinese information extraction system bring forth a solution for Chinese information extraction system carried out a valuable experiment carried out a valuable experiment still many works need to be done still many works need to be done lay a good foundation for our future works lay a good foundation for our future works
Thanks! Thanks! 谢谢! 谢谢!