Semantic Analyzer

Web application consists of 3 parts: text input panel, expert panel and analysis result panel.

A bit detailed description

Text Input Panel is the text field to input, edit text information and its further analysis.

After clicking on Add bookmark button of tabbed menu (Figure 3), the whole text will be saved in the data base. You can overlook and add saved text with Bookmarks button of tabbed menu (Figure 4).

Figure 3 Tabbed Menu:

Bookmarks – To add inputted text in the tabs

Add bookmark - To overlook tabs

Figure 4 Tabs in Semantic Analyzer

Experts

Experts panel is the table with the expert name in one column and its description in the other (Figure 5). Expert has stored knowledge base on the certain subject. It allows flexible text analyze using one or group of experts.

Figure 5 Experts panel

To make a text analyze, you should select an expert with a certain knowledge warehouse. Otherwise in the bottom right corner of the screen the following message will appear:

Figure 6 Information dialog

To select experts use «Select All» и «Deselect All» buttons. To select some experts from the table press and hold «Ctrl» button, and select by left mouse button necessary experts.

Analysis result.

Analysis result panel (Figure 7) contains semantic tree consisting of sub trees and nodes that reflects semantic relations in the text. A tree name defines the value of some part of the text. The node that does not come into any tree is considered not analyzed.

Figure 7 Analysis result panel

Start of work

To make an analysis of the text you should make expert check up. For example, creating an expert specialized in the dates (“dates expert”): let’s consider preparation process for the following analysis.

Press on the general menu button “To create a record”. Dialog window with “Name” and “Description” text fields appears. Let’s create the dates expert filling text fields as in the Figure 8. Save the record.

Figure 8 Expert creating

Teaching

Now we can teach the expert with knowledge that can help it to define the date and time in the text. You can use text input panel instruments on thus purpose (Figure 2). Copy the text that should be analyzed to the input field (Figure 9). Select just created dates expert in the expert table. Let’s teach the expert what month means. Select word “December” in the text and click on the selected word with left mouse button.

Fill text field Value in the appeared pull down menu as in the Figure 9 and press Enter. So we matched the word December to the word month. Then our analyzer can understand that December means month in the text.

Figure 9 Expert teaching

Text analysis

After teaching an expert of group of experts we can start analysis.

Input the required text and choose experts and press Analyze button. After completing analysis, the tree of text semantics will be built in the results panel.

So, using dates expert make sentence analysis:

«As of December 31, 2005, we employed approximately 1,300 full-time people, of whom over 400 were outside the United States.» (*)

Please consider that we have December month expert in our knowledge warehouse. Select dates experts and press on Analyze button. Here you are the analysis results in the figure 10.

Figure 10 Semantic analysis results

Teaching with mask

To make semantic analysis of definite word or group of words it is enough to teach expert using described above method. But in the every phrase or in the sentence the word is used in some context. For example, the word December can mean the month of the year or be the title of some café with delicious ice cream served in it.

To define the meaning of the phrase in the sentence, expert should be taught to understand semantic words relations and analyze their meanings.

The text is represented in the form of XML tags (Figure 11). The whole teaching text is described in the mask tag. Figure 6 means day in this context, and 1884 means year. These figures are described in the day and year tags respectively. Date value is set for this whole structure.

Lets make reanalysis of the studied above ( * ) sentence with a new knowledge warehouse. Its result can be seen in the Figure 12 that analyzer considered the 31 of December as the tree called “date” with month, day and year sub trees.

Knowledge warehouse

To study and edit expert knowledge warehouse you can use Edit button of the general menu (Figure 13). The dialog window with knowledge warehouse table will appear on the screen.

You can edit, add, delete and duplicate knowledge the same as experts (Figure 5).

Figure 13 Expert analysis

General knowledge

Knowledge warehouse can contain some knowledge of other experts as well. For example, we taught currency expert (Figure 14).

Figure14 Expert knowledge warehouse

Currency expert knowledge warehouse contains such notions as currency name, dollar as USA currency unit and dollar denotation. The top of the knowledge structure hierarchy is currency entity, and then currency name (currency_name) and its amount come; there is dollar under currency name. Besides dollar currency_name can contain Euro, Swiss franc, pounds sterling, etc.

In this case Currency Expert knows that there is such currency unit as dollar.

Therewith we have one more expert - Financial Data Expert.

Figure 15 Expert knowledge warehouse .

Let’s create one more expert and call it Financial Expert. Teach this expert to recognize a widespread phrase in the financial documents:

" total revenues of $2.02 billion".

Input this phrase in the text field and select Financial Expert in the table and make click on the left button of the mouse. Point total revenues in the Value text field of the pull down menu. Total_revenues entity appeared in the Financial Expert knowledge warehouse.