Natural language interface
(description of workflow)
The main advantage of this interface over traditional
menu-oriented interfaces is a possibility to retrieve the
information fast and efficiently by submitting a single phrase in
natural language instead of tedious selection of menu elements.
The natural language interface is flexible and adaptive both to
changes in the knowledge domain and user's views. It provides a
possibility to formulate arbitrary queries in various languages
(in English and in Russian, currently) to the FlyEx database.
To formulate and execute queries to the database the HTML form Natural Language Interface is
to be filled by a user (see Fig. 1).
Figure 1. Input query form of Natural Language Interface
The text of a query is entered in the text field QUERY. The list QUERY EXAMPLES contains
a set of predefined standard queries for convenience. On selection of a query from the
list this query is displayed automatically in the field QUERY and may be edited before execution.
By default a query returns all rows, which satisfy to selection criterion. The text field MAX. NUMBER OF
ROWS RETURNED allows to define a number of rows which will be returned as a result of a query.
If this number is smaller than the number of rows indicated in the field,
all rows will be returned.
The checkboxes "Search in" allow to specify the databases. It is possible to select any combination of databases. Note, that while both Mooshka and FlyEx provide
access to data on expression of segmentation genes in Drosophila blastoderm, they were developed using relational technology and technology of multidimensional databases, and differ in their contents.
Selection of the link SWITCH TO RUSSIAN calls the
Russian version of the query form. The queries in Russian
are submitted and executed similarly to queries in English
(see Fig. 2).
Figure 2. The Russian language version of the query form of Natural Language Interface
By pressing the button SEND QUERY a query will be
executed and after a while a result of the query will appear
in a new browser window. In the upper part of this window
the query in natural language is displayed, in which words
used to retrieve the information from the database are
shown in red. The SQL query, automatically generated by
the system, is presented below the NL query. SQL query
can be edited and returned to the server by pressing the
button SEND QUERY. The query result is displayed as a
table. Figure 3 presents the result of the query
"Which embryos were scanned for expression of Kruppel, giant and even-skipped?".
Figure 3. The result of the query "Which embryos were scanned for expression of Kruppel, giant and even-skipped?"
To formulate a NL query a user can use any concept
described in the conceptual scheme. Both queries which
use higher level or lower level concepts are interpreted by
the system equally well. For example, the cleavage cycles
from 11 to 13 are parts of developmental stage 4 and the
query `What embryos belong to stage 4?' returns a list of
embryos belonging to these cycles.
To formulate a NL query a user may type the words in
any word form (for example, embryo or embryos, gene or
genes). The query can be formulated both
as a whole phrase or as a list of keywords. For instance,
the query `embryos Kr gt eve' will return the same result
as the query `Which embryos were scanned for expression
of Kruppel, giant and even-skipped?'
The queries can be formulated using synonyms or even
laboratory jargon. For example, a gene name can be
introduced both as a full name or a symbol. To retrieve
the information about a gene group one can equally use
terms `gene group' and `combination of genes' or even
abbreviation in the form of three capital letters, each of
which corresponds to the common notation of the gene
(e.g. `BHE' - group scanned for expression of bcd, gt and eve).
The query `How many ...?' allows to count rows
satisfying any criterion (e.g., .How many embryos are
scanned for expression of bcd and belong to late temporal
classes?'(see see Fig. 4). Besides, it is possible to list these rows on a screen (for example, `How many fluorophores were used to detect expression of eve? List these fluorophores.'(see Fig. 5).
Figure 4. The result of the query "How many embryos are scanned for expression of bcd and belong to late temporal classes?"
Figure 5. The result of the query "How many embryos are scanned for expression of bcd and belong to late temporal classes? List these embryos."
One of the important features of the system consists
in a possibility of arbitrary assignment of conditions
posed on values of numerical attributes. The following
combinations of semantic constructions are supported:
larger than, greater than, more than, >, > =, <, less
than, smaller than, from n to m, n - m.
It is possible to combine selection criteria in a query
using logical operators AND, OR, NOT. For example, the
query `Which embryos were scanned for expression of
Kruppel and giant and even-skipped?' returns a list of
embryos, which were scanned for expression of all these
genes (see Fig. 3), while the query `Which embryos were scanned for
expression of Kruppel or giant or even-skipped?' returns
a list of embryos scanned for expression of at least one of
these genes (see Fig. 6).
Figure 6. The result of the query "Which embryos were scanned for expression of Kruppel or giant or even-skipped?"
The query `Display pattern ...' returns a pattern of
segmentation gene expression. When several patterns
in different embryos are requested, an embryo list is
displayed, in which embryo names are linked to embryo
images (see Fig. 7). If a pattern in one embryo is requested, it will be displayed at once, instead of the embryo list. If the pattern
of only one gene is requested a single stained image is displayed (see Fig. 8), otherwise a multiple
stained image is presented to the user (see Fig. 9).
Figure 7. The result of the query "Display patterns of expression of Kr in embryos belonging to temporal class 2."
Figure 8. The result of the query "Display patterns of expression of Kr in embryo cq7."
Figure 9. The result of the query "Patterns of expression of all genes scanned in embryo cq7."
Quantitative and processed expression data can be displayed to a user in different formats.
To retrieve this information a user has to specify a desired format of a query , e.g. `Select as a flat graph a quantitative data on expression of Kruppel in embryos belonging to temporal class 3'.
The quantitative gene expression data selected from the FlyEx database can be presented to the user as a table (see Fig. 10), flat graph or 3-D graph.
Figure 10. The result of the query "Display quantitative gene expression data for embryo cq7 as a table."
Registered gene expression data are presented to the user as a table or as a graph. FlyEx stores two sets of registered data obtained by two different registration methods: SpA or
FRDWT.
(
Go to the same page in Natural Language Interface
)
(
Go to the same page in Natural Language Interface
)
(
Go to the same page in Natural Language Interface
)
(
Go to the same page in Natural Language Interface
)
(
Go to the same page in Natural Language Interface
)
(
Go to the same page in Natural Language Interface
)
(
Go to the same page in Natural Language Interface
)
(
Go to the same page in Natural Language Interface
)
(
Go to the same page in Natural Language Interface
)
(
Go to the same page in Natural Language Interface
)