ajay154 / c-phrase Goto Github PK

Get support for the very basic temporal conditions

OK Björn seems to have enabled this type of capability in an earlier C-Phrase. 
Now it should be integrated into the main branch.

You need to make sure that schema is saved in topological sort order.

What are the steps to adding C-Phrase to an existing database that was not 
previously designed to work with C-Phrase?

The System Guide mentions that this is one of the intended uses of C-Phrase, 
but only documents using C-Phrase with the included demo "geo" database, which 
includes a pre-generated C-Phrase schema file. It's not clear to me how I would 
wrap C-Phrase around an arbitrary PostgreSQL database.

In guide.pdf, step 4 in section 2.2 "Installing the basic system" fails.

"Go to the directory C_PHRASE_ROOT/tests and type: make hash."

myuser@localhost:/tmp/c-phrase/tests$ which c-phrase
/usr/local/lib/c-phrase/bin/c-phrase
myuser@localhost:/tmp/c-phrase/tests$ make hash
make: *** No rule to make target `hash'.  Stop.

I'm running Ubuntu 10.04. 

In the long run, the queries must be able to express disjunction. This however, 
should be put off until the system is completely re-factored. 

It will be a fairly extensive enhancement and will touch practically 
everything. Yikes! 

When C-Phrase imports the schema of a new database, it should also import (as 
relations) all defined views over database.

Enhance answer strategy processing.

The config.ini is needlessly complex. Better to simply use cookies.

Put an extra space in an attribute name you add. 

Should be compressing strings...

What steps will reproduce the problem?
1. Start c-phrase server without a config file
2. Connect via admin interface
3. Do some work
4. Click save

What is the expected output? What do you see instead?

Should prohibit or ask user for name.
Instead a config file named NULL.cph is created.

It is annoying when one browses the ontology and can pick relation/attribute
combinations for which there are no entries. 

Should be easy to fix. 

I need to go through all the code, re-factor where needed and then 
finish commenting. This is fairly high priority at this stage. 

1. Trace: 
(PARAPHRASE '((X POPULATION) CITY (= X NAME "Austin") (= X STATE "Texas"))
 '(:DEF :SING) '(:PROPER-NAME))
1. Trace: PARAPHRASE ==> "population"

It is a case overlooked in raise-to-equiv-rewrite.

For a database with serial types, point C-Phrase via ODBC to database.
As C-Phrase brings over the initial schema it will interpret serial types
as integers.

What steps will reproduce the problem?
1. Define a concept using a value in an attribute position.
 Define a "jainitor" as a person with position jainitor
 (we assume that janitor is sourced)

2. Later try to ask a query using that condition
 How many persons with position jainitor?


What is the expected output? What do you see instead?

Well a tagging that let's string value "janitor" should be entertained.
But it looks like the tagging of "janitor" to a lex value dominates.

Should be easy to fix one I re-open the tagging stuff.

What steps will reproduce the problem?
1. Simply make any change through the admin interface.
2. Launch natural language query that takes advantage of change

What is the expected output? What do you see instead?

Admin changes should be reflected in the nl::*rules* list. They aren't. 

Try and build a schema with alter on. You will not be able to delare foreign 
keys. 

Currently the tagger does not correctly identify new names. It is limited to 
unknown spans of size one and, for example in the case of "Des Moines" it will 
identify "Des" and "Moines" as unknown spans. 

Should be relatively easy to correct and test. When I get around to it.  

Simply evaluate a query such as:

 (unique?  
'(X TAKES (+ (Y21) (COURSE Y21) (= X COURSE Y21 ID) (= Y21 NAME "ADMS"))
 (+ (Y11) (STUDENT Y11) (= X STUDENT Y11 ID) (= Y11 NAME "Mike")))

I really need to clean up/uniform the kernel... Yikes. 

A bug in the GUI to the substitution interface.

I'm unable to complete step 3 in section 2.3 of guide.py. I have PostgreSQL 
installed and running. I've created the "mike" user. I've logged in to the 
database using pgAdmin III, confirming the database exists and user's login 
works. I've populated <cphrase_root>/odbc.ini.

However, the instructions aren't clear on how to use odbctest. The FullConnect 
menu lists SQL_NULL_HANDLE for the environment handle. Entering "geo" for DSN, 
and then the username and password produces the error:

Diag(hdbc 0x1262ac0):IM002:0:[unixODBC][Driver Manager]Data source name not 
found, and no default driver specified
SQLConnect() failed

What else needs to be done to create the DSN for odbctest?

A bit surprising that I never got around to this. 
Any way, it should be easy. 

What steps will reproduce the problem?

Simply launch the interface, do some update and then observe that 
ERROR: invalid byte sequence for encoding "UTF8": 0xe46e6e;

Simply try to add a *Z into an edit on a pattern

Pick an relation/attribute value (e.g. STUDENT HEIGHT) in the Ontology pane.
Notice how superlatives top-height do NOT show up in the set of matching
concepts.

When author supply two join names through bordering in the demo, the
generator crashes.

Earlier C-Phrase could venture a report false presuppositions on values.

So if a user typed:

 what is the population of Gerpy? 

the system would respond:

 there is no state with name "Gerpy"
 there is no city with name "Gerpy"
 ...

This is a key usability feature that needs to be brought back.

Make a spelling error that is corrected and see that the interface does not
report error correction al la google.