- tech & impact
- components of a semantically enabled system
- tech uses?
- how might semantics affects your business?
Semantically Enabled Systems aim to understand the text being reviewed, in order to decide whether the text is relevant or not. It does this by analysing the words and sentence construction. These techniques allow us to find the information we are looking for, even though it may be in a different form than we were expecting. However, it is not as straight forward as it sounds.
- Morphology - Words are written differently depending on their tense (eg tells, telling, tell);
- Sentence structure – sentences appearing to be similar in structure often differ in analysis:
- John pulled the door with a bang (i.e. bang refers to the door)
- John pulled the door with a grin (i.e. grin refers to John)
- Word meaning – Words can have multiple meanings:
- bank (i.e. Noun a repository for money)
- bank (i.e. Noun meaning the edge of a river)
The context is necessary to determine which meaning is correct.
Semantically Enabled Systems use a variety to tools to deconstruct the text, identify named entities (eg people) and disambiguate the meaning of words given their context. They also have other key components:
- Ontologies – data structures that define entities/classes and their relationships to other entities. For instance, Teacher might be a type of a more general class of Worker.
- Dictionaries – to provide real world context
- Synsets - to provide other words with similar meanings
- Gazetteers - to provide context/domain specific vocabulary
- Rules Engines – artificial intelligence based programming environments to enable the development of code to differentiate or infer meaning.
The ability to understand the text being reviewed gives the opportunity to add new functionality to applications.
This could be:
Information Extraction
Means: Extracting, cross checking, standardizing information before storing
Uses: Application form processing, CV Processing, to trawl web logs and capture positive and negative references to your or competing products...
Relevance Checking
Means: Determining whether the content of a document is relevant
Uses: To monitor customer or regulator web sites for changes
Categorising
Means: Determining the subject matter of the document
Uses: To monitor RSS feeds for references to customers
Data Mining
Means: Finding important information in large volumes of text
Uses: To find references to competitors or customers
Document Distribution
Means: Routing documents dependant upon their content.
Uses: Routing Customer emails to appropriate service Agents
Semantically enabled systems will allow organizations to intelligently process information with less human interaction. This means provides an opportunity to reduce cost or process much more information for the same labour cost.
For instance:
- an organization that routinely (but manually) reviews competitor websites monthly - could automate that review process and run it daily, sensing the changes made and route the changes to the appropriate person;
- an organization processing orders or customer enquiries can automatically route emails to the appropriate department;
- an organisation manually reviewing customer feedback on 3rd party websites, can run the process daily or at will, in response to certain external events.
These systems will also allow organizations to offer something radical to their users:
- An organization wishing to place make its document library available on line is reliant upon its user base being able to guess the relevant free text search. Alternatively, they could:
- Automate the classification of each document;
- Automate the summary of each document
- Build a search engine based on synonyms
- Use ontologies to provide improved relativity between concepts
- Build a natural language based search engine
Whilst these are more difficult to conceive and conceptualise, they do offer bigger prizes, including the opportunity to change the dynamics of the industry.