 |
|
|
|
| REFLEXION: Search Functionalities |
|
These functionalities may be tested in the online demo.
They are presented in detail in the REFLEXION software documentation.
|
| Full text search |
REFLEXION is exhaustive.
It searches through the whole document system which has been indexed.
All the words are taken into account: those found in language dictionaries, and those which are not, i.e. new words and technical words.
|
| Queries with few restrictions |
The user may type into the query any combination of letters, numerals and other characters.
The number of keywords is not limited. However, to get the best results it is advisable not to enter more than 5 to 6 words.
REFLEXION recognizes special characters (accents, etc) according to the document language.
The joker '*' may be used to represent any string of characters ending a keyword.
|
To search by concepts:
the grammatical ("inflected") forms |
A given word can have various forms in its language: those forms are called inflected in linguistics, (we also call them "grammatical" forms): gender, singular/plural, verb conjugated forms, déclinations in some languages, etc... The word to which they pertain is called the "lemma".
REFLEXION can recognize the lemma from the particular form entered by the user in the query, so it yields the same results whatever the form.
For example in English, a query with ox or oxen will give the same results.
This functionalities, ignored by most Internet search engine, is a major key to obtain responses matching the concepts intended by the user.
|
| Treating natural languages |
REFLEXION is multilingual, it accepts all alphabétical languages.
The user may type his/her query in any of those languages, even if the different words pertain to different languages, so REFLEXION is cross-lingual.
REFLEXION retrieves the documents, in the different languages, which contain the query words.
For languages whose REFLEXION has a dictionary, it knows the lemmas of the key words, and retrieves documents which contain them. For other languages, searching is made with the entered forms, and is therefore less accurate but still very useful.
In the current version, REFLEXION holds linguistic dictionaries in English, French and German, each containing serveral hundreds of thousands of word forms. It automatically recognizes the dominant language of each document.
This ability of treating natural languages is a decisive factor of pertinence of a search engine dedicated to enterprises.
Note: this functionality does not equal to automatic language translation, but is a major step towards a "metalanguage" allowing concept retrieval independently of specific languages.
|
For a semantic searching:
phrases and words proximity |
The expression search option will give as results all documents which contain exactly the sequence or phrase entered in the query.
The distance option, in number of words, restricts the results to the documents where the first and the last words are separated by less than the chosen distance.
The proximity of words in a same document is a good clue of the pertinence of the given document regarding the set of concepts, or themes, which are of interest to the user.
|
| Sorting results by pertinence |
REFLEXION displays the list of results sorted by decreasing pertinence.
The pertinence ranking algorithm takes several factors, among which:
the relative weight of each document in the indexed set, itself weighted by language,
the number of keywords occurrences in each document and in the whole set of documents,
|
| Control the user's access to documents |
When receiving a query, the REFLEXION search engine knows the user's identity (authentication). Before displaying the list of results of a query, it checks for the access rights of that user to each document of the list, and displays only the documents allowed to the user (authorization).
Thus a user will only see the documents for which he/she has read rights
Controlling the users access to information is an essentisl capacity for an enterprise search engine.
|
| More search functions... |
Boolean operations (AND, OR, NOT): they are offered by REFLEXION in a very user fiendly form, without command language.
"Banal" or "stop" words: those are small words like conjunctions, pronouns, etc, wich do not represent objects, concepts neither actions. They are ind'exed but are skipped by REFLEXION at search time 'excepted when searching for a phrase). For each supported language, REFLEXION has a list of about 200 to 300 banal words that the Administrator may modify.
Filtering of search results by various parameters (language, date, document size, etc...
|
|
|
|
Solutions
