|
A language processing module is a learnable module that can understand and analyze language based on what it has learned.
 |
natural Semantic Modules develops a new generation of language processing modules.
Our modules are extremely specific in delivering the information required by a customer (precise answer to the question
as opposed to a selection of web links), and they are flexible and easily customizable. |
 |
The module has two basic usage modes that are the sources of internal knowledge and subsequent evaluation process:
- Learning mode
- User mode
allows the creation of relationships between sentences and desired functionality. The relationships are in the form of conceptually-specific synonymous phrases represented by a set of associations that describe each concept with relation to all other concepts. Domain-specific relationships allow the module to evaluate language with respect to the current application, thus reducing ambiguity and allowing for a deep understanding of intended meaning.
utilizes a set of powerful and flexible discovery rules allowing the module to combine contextual and conceptual understanding on the fly. The rules recognize important features of user input and generate dynamic components which represent a complete response path to the input, including how to a non-exact mode in the case that an exact answer is not found. These discovery rules seamlessly use basic concept definitions and acquired relationships in both the understanding of linguistic features of user input and the generation of the response. To further respond to user request, the rules could accept input from third-party repositories (such as CRM systems).
When content or user input is processed through the module, a series of sophisticated linguistic and statistical techniques are applied, each of which builds on previous steps to add to the complete understanding of the content or question currently being evaluated. Once the user input is fully processed, the module produces a set of relationships which are concurrently evaluated against the content sources present to retrieve an exact answer to the input. If an exact answer doesn't exist (or, in the case of general questions, a non-specific answer is more appropriate), the module automatically backs off of the strict requirements of the initial processing to look for the most appropriate solution in a non-exact mode. This allows the module to automatically retrieve the best answer available in the content sources present, as well as alert content owners to exact cases where more detail might be necessary.
|
