Automatically structuring domain knowledge from text: An overview of current research

Question

1. What are the contributions in "Automatically structuring domain knowledge from text: a review of current research" ?

2. What future works have the authors mentioned in the paper "Automatically structuring domain knowledge from text: a review of current research" ?

3. What is the main disadvantage of unsupervised learning methods?

4. What is the main disadvantage of weakly supervised learning methods?

Accepted Answer

This paper presents an overview of automatic methods for building domain knowledge structures ( domain models ) from text collections.. The authors will also discuss trade-offs between different approaches and point to some recent trends.

Accepted Answer

The scope for further research in this area looks promising, given the growing number of popular search engines that have started employing optional interactive visualisations of term relationships.

Accepted Answer

Unsupervised learning can be applied to the actual textual data sources to build a domain model or to implicit data sources such as query logs and click information associated with text collections.

Accepted Answer

Weakly supervised learning methods could benefit from research oriented towards broadening the coverage of relationship types in a way (for example, active learning) that actively selects new relationship types with respect to a selected application or domain.

Accepted Answer

Self-organising maps have also been applied to enrich the relationships between concepts (Dittenbach et al. [2004], Chen et al. [2008]).

Accepted Answer

Data-driven domain modelling approaches have been widely used in IR and NLP, which are examples of research fields that have seen a shift from mainly symbolic ideas to a strong preference for the statistical approaches.

Accepted Answer

The advantage of weakly supervised learning methods of domain modelling is that explicit concepts and relationships become available through targeted harvest.

Accepted Answer

The advantage of supervised learning methods is that the annotation of training data can be carried out in such a way that all relationships and concepts are selected to meet the needs of the selected domain or application.

Accepted Answer

The advantage of unsupervised methods is that little human labour is required to producewell annotated training data, but the main drawback is the difficulty in producing annotations of explicit concept classes and relationships.

Accepted Answer

The reliance of applications on well-designed data structure leads the research in this community to be largely dominated by semi-automatic and manual approaches (e.g., Flouris et al. [2008], Maedche et al. [2003]).

Accepted Answer

In particular, the ability to extract formal terminology and identify various types of semantic relationships between the terms (a.k.a. ontology) from unstructured text is considered to be of critical importance (Navigli and Velardi [2008], Buitelaar et al. [2005]).

Accepted Answer

The relationships are often extracted using co-occurrence frequency within the collection or using inferred attributes of the concepts.

Accepted Answer

In particular, the extraction of concepts from text followed by an analysis of cooccurrence statistics (that is the counts of two concepts occurring within close proximity within selected text) as an approach to information searchand seeking was already being mentioned by Doyle [1961] at a time when computational resources were limited.

Accepted Answer

Examples include adapting domain models based on user feedback on the relevance of documents (that is, no explicit judgement on the domain model itself) (Nanas [2003]) as well as adaptive domain models that learn from user query modifications in interactive search (Kruschwitz et al. [2011]).

Accepted Answer

Two ontologies covering the same domain may use different concept and relation designators in different configurations, which makes similarity comparisons difficult.

Accepted Answer

The patterns induced, within this framework, can have constraints not only on surface patterns such as lexicon and part-of-speech, but also constraints on the semantic classes of the words in the pattern.

Accepted Answer

bipartite graphs that include both queries and URLs as nodes can be used to identify a domain model of closely related terminology, that is, phrases that have resulted in the retrieval of the same documents (Deng et al. [2009], Craswell and Szummer [2007]).

Accepted Answer

Some tools, however, such as the Karlsruhe Ontology (KAON) framework11 and OntoLearn (Navigli et al. [2004]), actively support language processing for automatically extracting and selecting keywords representative of domain concepts from natural language texts.

Automatically structuring domain knowledge from text: An overview of current research

Chat with Paper

AI Agents for this Paper

Most frequently asked questions

1. What are the contributions in "Automatically structuring domain knowledge from text: a review of current research" ?

2. What future works have the authors mentioned in the paper "Automatically structuring domain knowledge from text: a review of current research" ?

3. What is the main disadvantage of unsupervised learning methods?

4. What is the main disadvantage of weakly supervised learning methods?

5. What has been used to enrich the relationships between concepts?

6. What are the main reasons for the use of data-driven domain modelling?

7. What is the advantage of weakly supervised learning methods of domain modelling?

8. What is the advantage of supervised learning methods?

9. What is the main drawback of unsupervised methods?

10. What is the main reason for the reliance of applications on well-designed data structure?

11. What is the importance of identifying semantic relationships between terms?

12. What is the common way to extract relationships?

13. What was the main disadvantage of the extraction of concepts from text?

14. What are some examples of adaptive domain models?

15. What makes the comparison of ontologies difficult?

16. What are the constraints on the semantic classes of the words in the pattern?

17. What is the main idea behind the use of bipartite graphs?

18. What tools are used to extract and select keywords representative of domain concepts from natural language texts?

Figures

Citations

Modeling Relation Paths for Knowledge Graph Completion

Looking for Information -- A Survey of Research on Information Seeking, Needs, and Behavior by Donald Case

Identifying Fake News on Social Networks Based on Natural Language Processing: Trends and Challenges

Formal Concept Analysis – Overview and Applications

Searching the Enterprise

References

WordNet : an electronic lexical database

DBpedia: a nucleus for a web of open data

Yago: a core of semantic knowledge

Automatic acquisition of hyponyms from large text corpora

Distant supervision for relation extraction without labeled data

Related Papers (5)

Exploiting online human knowledge in Requirements Engineering

The Semantic Snake Charmer Search Engine: A Tool to Facilitate Data Science in High-tech Industry Domains

OECEP: enriching complex event processing with domain knowledge from ontologies

Semantic Web Requirements through Web Mining Techniques

An Introduction to Semantic Web Technologies