Overview

HOLTRAN Patented Technology

builds a bridge over the existing precipice between the two fundamental methods of representation of information: unstructured texts and structured data. On one brink we have documents in natural languages being capable to express arbitrary content that can be percept only by human. On the other - we see   various machine oriented data models, each of which is capable to keep only certain content and generally is incompatible with others. HOLTRAN Technology allows managing effectively both kinds of data.

Use Cases

Financial traders need fast, simple and reliable interfaces to pricing and trading systems capable of translating their cryptic financial jargons to standard compliant formats. HOLTRAN technology provides such capability due to its universal representation of syntax-to-semantics interface, allowing freely switching between different input and output languages to access the same semantic content. Such flexibility means that the trader can copy and paste a received quote into the command line and tanslate it into the database or the calculation engine query.

We provide two financial applications:

FJ (Financial Jargon) translator capable to translate a free format description of a financial instrument to Internet pricing system schemes such like generic XML, RDF, FpML or FIXatdl.

BQL (Blotter Query Language) translator capable to translate a free format query to a data base of trade records into XML, RDF or OWL formats (eventually to be converted to SQL).

Assume there is a set of documents on various aspects of activities of automative industry companies. In fact, the Internet or an enterprise Intranet may also be considered as such sets, though distributed between multiple sites and, of course, joined with a lot of documents on other subjects.

The regular search engines can pretty efficiently find the documents on the interesting subject out of the entire Internet, but the analysis of the contents of the found documents they still leave to the user.

Now assume we need to find out from these documents an answer to a simple question like: "What are the partners of GMC?". Given such a question, HOLTRAN system will automatically extract from each of the available documents precise and detailed information related to the question and merge it into a single and complete answer, that is just a list of all GMC partners. It also might (optionally) include into the answer references to concrete documents from which certain information was extracted.

Assume that the answer to the above question mentions, in particular, Toyota, Fiat and also a long list of unknown to us companies, most likely GMC dealers. New, more complicated questions may be then addressed to the system, for example: "With which of its partner does GMC have common projects for new engines design?" or "Which of GMC dealers made the highest volume of sales in Europe last year?” To answer these questions the system should retrieve from the available documents more specific details on GMC partnership as well as on all individual GMC dealers, analyze them and compose appropriate answers.

A distinctive feature of the HOLTRAN technology consists in the fact that it converts textual documents to be analyzed into the higher-order relational database (to which we reference as to HOLTRAN knowledge base) in order to establish content links and thus enable meaning-based (rather than text-based) search and analysis.

In the above example, the actual conversion of the documents may be conducted either in the process of their retrieval via the network or preliminarily, upon submission of documents to the enterprise information system. In the latter case the processing of end-user's requests will be, of course, much faster.

Consider another example concerning the book publishing. Having the two vaults, one containing a log of customer orders and the other - abstracts of the books (and/or articles, say scientific ones), we would likely be able to compare their contents in order to determine which of newly issued books or articles should be advertised to which of customers. A HOLTRAN-based application might fulfill this task by extracting from the abstracts and taking into consideration such characteristics like theme, direction and/or point of view of author, grade of complexity, style of exposition etc.

The most important is that the HOLTRAN technology allows not only to flexibly combine and prioritize (depending on specific company requirements) some pre-defined characteristics, but also to easily expand and enrich them. For example, a publishing company specialized in historical books might want to additionally take into consideration such a characteristic like whether a book describes real historical events and personalities or presents a kind of "historical fiction".

Note that in order to add such a new characteristic to the list of ones considered by the system there is no need in either re-working neither re-submitting the existing book abstracts to the system: it is enough only to define a new concept, that is to "explain" it to the system basing on (referencing to) previously defined ones.

This explanation may be formulated in a pseudo-natural language by an experienced end-user. Of course, in order to effectively "understand" new explanations the system (more specifically - the HOLTRAN knowledge base) should possess a serious basic knowledge in the corresponding area.

Establishing the basic knowledge is also performed by subsequently defining more complicated concepts from simpler ones. Thus, the development of a HOLTRAN application recalls teaching and training rather than regular software programming.

Moreover, if at a certain stage, we decide the system to take into account also the readers' surveys and feedbacks we do it principally in the same way as above, i.e. without any additional software development.