diff options
| -rwxr-xr-x | python/extraction_cdx_grobid.py | 8 | ||||
| -rwxr-xr-x | python/extraction_ungrobided.py | 8 | ||||
| -rw-r--r-- | scalding/src/main/resources/slug-denylist.txt (renamed from scalding/src/main/resources/slug-blacklist.txt) | 0 | ||||
| -rw-r--r-- | scalding/src/main/scala/sandcrawler/ScorableFeatures.scala | 6 | ||||
| -rw-r--r-- | scalding/src/test/scala/sandcrawler/ScorableFeaturesTest.scala | 2 |
5 files changed, 12 insertions, 12 deletions
diff --git a/python/extraction_cdx_grobid.py b/python/extraction_cdx_grobid.py index 040538c..b27d053 100755 --- a/python/extraction_cdx_grobid.py +++ b/python/extraction_cdx_grobid.py @@ -37,7 +37,7 @@ from grobid2json import teixml2json sentry_client = raven.Client() # Specific poison-pill rows we should skip -KEY_BLACKLIST = ( +KEY_DENYLIST = ( 'sha1:DLCCSMMVTCCIR6LRXHEQLZ4PWO6NG2YT', # "failed to guess ARC header format" ) @@ -210,9 +210,9 @@ class MRExtractCdxGrobid(MRJob): yield _, status return key = info['key'] - if key in KEY_BLACKLIST: - self.increment_counter('lines', 'blacklist') - yield _, dict(status='blacklist', key=key) + if key in KEY_DENYLIST: + self.increment_counter('lines', 'denylist') + yield _, dict(status='denylist', key=key) return # Note: this may not get "cleared" correctly diff --git a/python/extraction_ungrobided.py b/python/extraction_ungrobided.py index 4074112..972b8f9 100755 --- a/python/extraction_ungrobided.py +++ b/python/extraction_ungrobided.py @@ -37,7 +37,7 @@ from grobid2json import teixml2json sentry_client = raven.Client() # Specific poison-pill rows we should skip -KEY_BLACKLIST = ( +KEY_DENYLIST = ( 'sha1:DLCCSMMVTCCIR6LRXHEQLZ4PWO6NG2YT', # "failed to guess ARC header format" ) @@ -212,9 +212,9 @@ class MRExtractUnGrobided(MRJob): yield _, status return key = info['key'] - if key in KEY_BLACKLIST: - self.increment_counter('lines', 'blacklist') - yield _, dict(status='blacklist', key=key) + if key in KEY_DENYLIST: + self.increment_counter('lines', 'denylist') + yield _, dict(status='denylist', key=key) return # Note: this may not get "cleared" correctly diff --git a/scalding/src/main/resources/slug-blacklist.txt b/scalding/src/main/resources/slug-denylist.txt index 926dbd5..926dbd5 100644 --- a/scalding/src/main/resources/slug-blacklist.txt +++ b/scalding/src/main/resources/slug-denylist.txt diff --git a/scalding/src/main/scala/sandcrawler/ScorableFeatures.scala b/scalding/src/main/scala/sandcrawler/ScorableFeatures.scala index be2b495..95a39aa 100644 --- a/scalding/src/main/scala/sandcrawler/ScorableFeatures.scala +++ b/scalding/src/main/scala/sandcrawler/ScorableFeatures.scala @@ -8,8 +8,8 @@ import scala.util.parsing.json.JSONObject object ScorableFeatures { // TODO: Add exception handling. - val fileStream : InputStream = getClass.getResourceAsStream("/slug-blacklist.txt") - val SlugBlacklist : Set[String] = Source.fromInputStream(fileStream).getLines.toSet + val fileStream : InputStream = getClass.getResourceAsStream("/slug-denylist.txt") + val SlugDenylist : Set[String] = Source.fromInputStream(fileStream).getLines.toSet fileStream.close val MinSlugLength = 8 @@ -44,7 +44,7 @@ class ScorableFeatures private(title : String, authors : List[Any] = List(), yea val slug = StringUtilities.removePunctuation((unaccented.toLowerCase())).replaceAll("\\s", "") if (slug.isEmpty || slug == null - || (ScorableFeatures.SlugBlacklist contains slug) + || (ScorableFeatures.SlugDenylist contains slug) || (slug.length < ScorableFeatures.MinSlugLength)) { None } else { diff --git a/scalding/src/test/scala/sandcrawler/ScorableFeaturesTest.scala b/scalding/src/test/scala/sandcrawler/ScorableFeaturesTest.scala index 112a5e5..c847296 100644 --- a/scalding/src/test/scala/sandcrawler/ScorableFeaturesTest.scala +++ b/scalding/src/test/scala/sandcrawler/ScorableFeaturesTest.scala @@ -42,7 +42,7 @@ class ScorableFeaturesTest extends FlatSpec with Matchers { it should "filter stub titles" in { titleToSlug("abstract") shouldBe (None) titleToSlug("title!") shouldBe (None) - titleToSlug("a real title which is not on blacklist") shouldBe Some("arealtitlewhichisnotonblacklist") + titleToSlug("a real title which is not on denylist") shouldBe Some("arealtitlewhichisnotondenylist") } it should "strip special characters" in { |