Smarter popcat

[that1guy]

Macbook air popcat query

http://localhost:4043/v1/groupings/popular?query=macbook+air

Response shows vehicles and rentals as popular category

Think this is caused because popcat is searching body text (we should probably only search heading). Also, popcat is searching for matches of word "air" and getting things like "air filter" for cars.

[
  {
    "count": 2028,
    "code": "VAUT"
  },
  {
    "count": 906,
    "code": "RHFR"
  },
  {
    "count": 580,
    "code": "SELE"
  },
  {
    "count": 561,
    "code": "VPAR"
  },
  {
    "count": 407,
    "code": "SAPP"
  },
  {
    "count": 296,
    "code": "SAPL"
  },
  {
    "count": 120,
    "code": "VMOT"
  },
  {
    "count": 112,
    "code": "SOTH"
  },
  {
    "count": 101,
    "code": "STOO"
  },
  {
    "count": 98,
    "code": "RHFS"
  },
  {
    "count": 91,
    "code": "SFUR"
  },
  {
    "count": 81,
    "code": "SHNG"
  },
  {
    "count": 77,
    "code": "RCRE"
  },
  {
    "count": 50,
    "code": "RVAC"
  },
  {
    "count": 45,
    "code": "SSNF"
  },
  {
    "count": 44,
    "code": "SBIK"
  },
  {
    "count": 40,
    "code": "SKID"
  },
  {
    "count": 38,
    "code": "SBIZ"
  },
  {
    "count": 34,
    "code": "SWNT"
  },
  {
    "count": 29,
    "code": "VMPT"
  },
  {
    "count": 29,
    "code": "RSHR"
  },
  {
    "count": 26,
    "code": "SIND"
  },
  {
    "count": 24,
    "code": "VOTH"
  },
  {
    "count": 24,
    "code": "SCOL"
  },
  {
    "count": 22,
    "code": "STOY"
  },
  {
    "count": 21,
    "code": "SHNB"
  },
  {
    "count": 14,
    "code": "RSUB"
  },
  {
    "count": 14,
    "code": "SBAR"
  },
  {
    "count": 13,
    "code": "SGAR"
  },
  {
    "count": 13,
    "code": "STVL"
  },
  {
    "count": 11,
    "code": "SANC"
  },
  {
    "count": 10,
    "code": "SMUS"
  },
  {
    "count": 8,
    "code": "SANT"
  },
  {
    "count": 5,
    "code": "SMNM"
  },
  {
    "count": 4,
    "code": "JTRA"
  },
  {
    "count": 4,
    "code": "RWNT"
  },
  {
    "count": 4,
    "code": "JFNB"
  },
  {
    "count": 4,
    "code": "JGIG"
  },
  {
    "count": 3,
    "code": "SJWL"
  },
  {
    "count": 3,
    "code": "SLIT"
  },
  {
    "count": 3,
    "code": "JSKL"
  },
  {
    "count": 2,
    "code": "APET"
  },
  {
    "count": 2,
    "code": "JWEB"
  },
  {
    "count": 2,
    "code": "JADM"
  },
  {
    "count": 2,
    "code": "PMSW"
  },
  {
    "count": 1,
    "code": "JEDU"
  },
  {
    "count": 1,
    "code": "SVCH"
  },
  {
    "count": 1,
    "code": "JENG"
  },
  {
    "count": 1,
    "code": "JFIN"
  },
  {
    "count": 1,
    "code": "STIX"
  },
  {
    "count": 1,
    "code": "PWSM"
  },
  {
    "count": 1,
    "code": "SVCP"
  },
  {
    "count": 1,
    "code": "MMSG"
  }
]

The future of popcat (not necessary for beta launch obviously).

Eventually it would be awesome if popcat searched for "bicycle" and "bike" via Elasticsearch using the query below.

http://localhost:4043/v1/groupings/popular?query=bike

[that1guy]

@brozeph This one starting to stand out a bit. When selling anything in sell box using two words we get odd suggestions like autos instead of appliances and furniture.

I'm selling my #washing machine OR #king size bed

[that1guy]

Expanding on this. As mentioned before: "due to the sheer amount of rentals we ingest from CL our popcat algorithm favors RHFR category all too often. We need to level the playing field by calculating confidence percentages in each cat."

[brozeph]

@that1guy - assigning to you to evaluate... the results appear 100x more accurate and the damn thing is lightning fast now (HUGE performance improvement). The output is identical as before but because the way in which the search works is so drastically different, it's worth evaluating.

[that1guy]

effin' awesome.. I see the build occurring now.. Can't wait to test. Killer stuff dude.

[that1guy]

How the fuck did you do that???! Dude.. I'm blown away.. I just giggled to myself.

[that1guy]

Wow man.. it's so accurate too.. this is scary cool.

[that1guy]

Closing.. good stuff dude.