abathologist / borgmanities Goto Github PK

At the moment, the paths are dependent on the current working directory.

• Eliminate "Random Selections". Belongs rather in PaperTitleGen.IO.
• PaperTitleGen.IO is written in a primitive style, reflecting the current state of my knowledge. Substantial gains in clarity can be made once I push to the next level, so
• Attain next level.

Reasons for this change:

1. I want to get familiar with parsing in Haskell.
2. In order to really refine the output, we need to have more advanced parsing of the text. Immediate issues that call for this include:
   • Stripping out all links (requires parsing out relatively complex string patters).
   • Rejecting sequences that contain tweet-speek ("RT", "u r", etc).
   • Rejecting phrases with terms in the black list.
3. Thus, we'll either need to do more parsing on the python side or the Haskell side. Given (1) and the fact that I don't enjoy working in Python---Point made.

Todo

• Modify python to return un-filtered unicode text of tweets, one per line.
• Parse this text in the Haskell.

Cf. https://docs.python.org/2/howto/unicode.html

Note: I ran into no problems running BotticelliBot in iTerm, but when I ran it in emacs-shell, the encoding caused problems. I guess the former sets the environment to handle unicode encoding to some extent, though I've also noticed some artifacts in the strings coming fin from python.

In PapterTitleGen.Gen.hs: Select, manicure, and assemble data from fields of TitleParts input.
The main function for this should be generateTitle (rename if desired).

I'm putting you as principle on this, @KitLiterate, but not to be burdensome! Mainly just to explore the utility of the issue tracker as a project-planning tool.

Change or add what you will.

• Only take definitions before “with” and other prepositions?
• Only take definitions up to “having” and coordinating particles (“or” “and”) too ?
• Remove phrases in parens that appear in definitions.

On a recent test I got back "manorial based jobs xlyqfwri rt dsw worldwide the fight" as a compliment. But, if we could catch plurals, the complement would have been "manorial based jobs", which would have worked nicely.

• Come up with idea for fix.
• Fix.

• Implement posting cli interface in twitter.py
• Wrap call to twitter.py -t "tweet" in Twitter.hs

WordsApi suddenly quit working, I think. I don't know if this is because I exceeded the daily allowance, or because of an issue on their end. But, in either case, they are terminating their "sandbox" services come February 20th, and will only be serving their api via Mashape (https://www.mashape.com/wordsapi/wordsapi), so I'll have to make this switch eventually. The fact that I only think WordsApi is the problem indicates serious shortcoming in the design of the Web/WordsApi.hs: I have no logging or error reporting here.

• Add logging to Web/WordsApi.hs
• Set WordsApi.hs to just use System.Process to interface with curl instead of Network.Curl.
• Implement the Mashape token and query protocol via curl commands in WordsApi.hs.

• Compile blacklist.
• Include crap words amongst the vile ones (like "http")—words that are nouns but suck for our puroses.
• Implement censor of selected nouns based on blacklist

For the purposes of refining the algorithm and monitoring its efficiency (namely in terms of how many times it has to restart due to failure to find fitting candidate data), logging seems essential. Also, it's just a good thing to learn to implement.

• Don’t let single letters count as nouns (‘r’ and ‘s’ count as a nouns because they name the letter, I guess).
• Eliminate all numbers

Each successful seedNoun should be recorded in a list of seedNouns, and these should not be repeated (at least not until we have extended the range of inputs available to the bot).

• enter seedNouns of successful title generation into src/resources/seedNouns.txt.
• check the selection of new seedNouns against seedNouns.txt.

If title generation fails > 100 times, end the program.

BotticelliBot has been exceeding its free month quota. This month, it did so in a mere 15 days. That's too pricey for a novelty project. WordNet has a Prolog version of its 3.0 database, and I love Prolog; plus, it would be better if it were running independently any how. Thus...

• Implement Prolog front-end program for performing queries.
  • Create modular interface for wn database files.
  • JIT loading of database files (i.e., database files won't load unless you actually execute a goal that requires one of their predicates.
  • Implement flexible, general purpose query system for getting word information.
  • Format query results as Json
  • Implement executable PrologScript interface to call program from CLI.
• Link Haskell to Prolog program.

It's currently linked with my account, via the user tokens.

I think tweets are running together, leading to some nonsense...