Put​ ​context​ ​at​ ​the​ ​heart​ ​of​ ​social​ ​media​ ​sentiment​ ​analysis

Social​ ​sentiment​ ​analysis today

Sometimes​ ​called​ ​opinion​ ​mining,​ sentiment analysis ​is​ ​the​ ​holy​ ​grail​ –​ ​the​ ​most effective​ ​way​ ​to​ ​find​ ​out​ ​what​ ​consumers​ ​think​ ​about​ ​a​ ​brand,​ ​product​ ​or​ ​event,​ ​in​ ​real time.​ ​It​ ​aims​ ​systematically​ ​to​ ​identify​ ​opinions​ ​in​ ​a​ ​document​ ​and​ ​give​ ​it​ ​a​ ​score​ ​on​ ​a scale​ ​of​ ​negative​ ​to​ ​positive.​ ​​ ​Marketers​ ​use​ ​instant​ ​sentiment​ ​data​ ​to​ ​make​ ​campaign adjustments​ ​in​ ​close​ ​to​ ​real-time,​ ​instantly​ ​feeding​ ​negative​ ​reviews​ ​to​ ​product​ ​teams and​ ​solving​ ​major​ ​customer​ ​issues​ ​more​ ​efficiently.There​ ​are​ ​two​ ​main​ ​approaches​ ​in use​ ​today:​ ​sentiment analysis​ ​based​ ​on​ ​keyword​ ​scoring,​ ​and​ ​a​ ​calculation​ ​based​ ​on predefined​ ​categories.

Keyword​ ​scoring​ ​means​ ​you​ ​give​ ​the​ ​word​ ​‘good’​ ​a​ ​positive​ ​score,​ ​the​ ​word​ ​‘bad’ ​a negative​ ​score.​ ​This​ ​approach​ ​has​ ​inherent​ ​flaws​ ​which​ ​some​ ​analysts​ ​try​ ​to​ ​fix​ ​by applying​ ​rules​ ​on​ ​top​ ​of​ ​the​ ​keyword​ ​scoring,​ ​but​ ​the​ ​underlying​ ​problem​ ​remains: keyword​ ​scoring​ ​only​ ​captures​ ​fragments​ ​of​ ​a​ ​message​ ​and​ ​struggles​ ​to​ ​accurately judge​ ​true​ ​meaning.​ ​Keyword​ ​scoring​ ​can​ ​deliver​ ​some​ ​relevant​ ​results,​ ​if​ ​focused​ ​on broad​ ​trends​ ​or​ ​applied​ ​at​ ​scale.​ ​But,​ ​at​ ​a​ ​more​ ​detailed​ ​level,​ ​a​ ​sarcastic​ ​tweet​ ​or​ ​a 13-year-old​ ​calling​ ​a​ ​new​ ​video​ ​game​ ​‘sick’​ ​brings​ ​the​ ​system​ ​to​ ​its​ ​knees.​ ​Because keyword-based​ ​sentiment​ ​technology​ ​can’t​ ​understand​ ​context​ –​ ​only​ ​individual​ ​words​ ​or small​ ​phrases​ ​–​ ​proven​ ​accuracy​ ​levels​ ​generally​ ​range​ ​between​ ​50​ ​to​ ​80%.​ ​This​ ​has led​ ​to​ ​many​ ​brands​ ​sidelining​ ​sentiment​ ​data​ ​in​ ​reports.

The​ ​second​ ​approach​ ​is​ ​based​ ​on​ ​the​ ​idea​ ​that​ ​you​ ​let​ ​the​ ​user​ ​categorise​ ​a​ ​few​ ​dozen results​ ​as​ ​a training​ ​set, ​and​ ​then​ ​let​ ​an​ ​algorithm​ ​use​ ​this​ ​to​ ​make​ ​decisions​ ​for​ ​future results.​ ​Accuracy​ ​for​ ​predefined​ ​categories​ ​is​ ​usually​ ​higher​ ​than​ ​for​ ​keyword​ ​scoring, but​ ​it’s​ ​far​ ​from​ ​perfect.​ ​This​ ​manual​ ​categorisation​ ​requires​ ​a​ ​huge​ ​time​ ​investment​ ​and an​ ​understanding​ ​of​ ​all​ ​topics​ ​that​ ​could​ ​potentially​ ​be​ ​linked​ ​to​ ​your​ ​brand.​ ​Because this​ ​method​ ​works​ ​with​ ​such​ ​narrow​ ​parameters​ ​for​ ​qualifying​ ​results,​ ​it​ ​typically​ ​yields fewer​ ​results​ ​overall.

In​ ​a​ ​third​ ​approach,​ ​some​ ​analysts​ ​prefer​ ​to​ ​bypass​ ​technology​ ​altogether​ ​and​ ​require people​ ​to​ ​code​ ​sentiment.​ ​The​ ​upside​ ​here​ ​is​ ​that​ ​accuracy​ ​is​ ​no​ ​longer​ ​a​ ​problem. People​ ​will​ ​disagree​ ​here​ ​and​ ​there​ ​(language​ ​is​ ​not​ ​an​ ​exact​ ​science),​ ​but​ ​overall results​ ​are​ ​excellent.​ ​The​ ​downside?​ ​On​ ​average,​ ​a​ ​person​ ​can​ ​only​ ​classify​ ​about​ ​100 documents​ ​per​ ​hour.​ ​Brands​ ​can​ ​get​ ​thousands,​ ​even​ ​tens​ ​of​ ​thousands,​ ​of​ ​mentions​ ​a day.

Beyond​ ​issues​ ​of​ ​coding​ ​consistency,​ ​by​ ​the​ ​time​ ​coders​ ​find​ ​the​ ​critical​ ​mention​ ​that matters,​ ​it​ ​might​ ​be​ ​too​ ​late​ ​to​ ​act.​ ​Even​ ​worse,​ ​if​ ​brands​ ​want​ ​to​ ​determine​ ​consumer sentiment​ ​around​ ​certain​ ​products​ ​or​ ​services​ ​for​ ​market​ ​research​ ​or​ ​collect​ ​opinions around​ ​a​ ​trending​ ​topic,​ ​the​ ​substantial​ ​costs​ ​involved​ ​in​ ​securing​ ​timely​ ​results​ ​make this​ ​a​ ​non-starter​ ​for​ ​many​ ​brands.

Technology and sentiment analysis: change​ ​is​ ​in​ ​the​ ​air

For​ ​many​ ​years,​ ​the​ ​industry​ ​has​ ​tried​ ​to​ ​develop​ ​a​ ​social​ ​sentiment​ ​technology​ ​that would​ ​provide​ ​the​ ​accuracy​ ​and​ ​level​ ​of​ ​automation​ ​needed​ ​to​ ​make​ ​sentiment​ ​analysis effective.​ ​Its​ ​time​ ​has​ ​finally​ ​come.

In​ ​a​ ​world​ ​where​ ​customer​ ​centricity​ ​is​ ​increasingly​ ​recognised​ ​as​ ​a​ ​vital​ ​competitive weapon,​ ​sentiment​ ​analysis​ ​has​ ​never​ ​been​ ​more​ ​important​ ​to​ ​get​ ​right.​ ​​Talkwalker​’s​ ​AI technology​ ​based​ ​on​ ​‘deep​ ​learning’ ​–​ ​an​ ​advanced​ ​method​ ​of​ ​training​ ​machines​ ​to learn​ ​–​ ​now​ ​lets​ ​brands​ ​capture​ ​customer​ ​sentiment​ ​with​ ​90%​ ​accuracy.​ ​For​ ​the​ ​first time,​ ​we​ ​can​ ​truly​ ​understand​ ​the​ ​meaning​ ​of​ ​full​ ​sentences.​ ​We​ ​can​ ​accurately determine​ ​customer​ ​attitudes​ ​and​ ​contextual​ ​reactions​ ​in​ ​tweets,​ ​posts​ ​and​ ​articles. Brands​ ​can​ ​use​ ​and​ ​cross​ ​sentiment​ ​indicators​ ​with​ ​a​ ​variety​ ​of​ ​data​ ​to​ ​drive​ ​a​ ​better understanding​ ​of​ ​what​ ​their​ ​customers​ ​are​ ​thinking.​ ​Teams​ ​can​ ​cut​ ​down​ ​on​ ​reaction time.​ ​​Critical​ ​posts​ ​can​ ​be​ ​detected​ ​and​ ​flagged​ ​immediately,​ ​in​ ​real​ ​time.

Looking​ ​beyond​ ​the​ ​post​ ​level,​ ​sentiment​ ​analysis​ ​unlocks​ ​improvements​ ​in​ ​the customer​ ​experience.​ ​In​ ​a​ ​classic​ ​example,​ ​a​ ​client​ ​of​ ​ours​ ​in​ ​the​ ​budget​ ​hotel​ ​sector noticed​ ​a​ ​stream​ ​of​ ​negative​ ​posts​ ​on​ ​social​ ​media.​ ​They​ ​ceased​ ​when​ ​hair​ ​dryers​ ​were installed​ ​across​ ​all​ ​room​ ​categories.

Another​ ​client,​ ​a​ ​household​ ​brand,​ ​evaluated​ ​attitudes​ ​towards​ ​kitchen​ ​cleaners​ ​on​ ​the market,​ ​and​ ​was​ ​able​ ​to​ ​see​ ​that​ ​smell​ ​was​ ​one​ ​of​ ​the​ ​most​ ​commonly​ ​disliked​ ​issues. Based​ ​on​ ​this​ ​research,​ ​they​ ​asked​ ​their​ ​R&D​ ​team​ ​to​ ​come​ ​up​ ​with​ ​a​ ​product​ ​they could​ ​market​ ​with​ ​‘great​ ​smell’​ ​as​ ​a​ ​differentiator.

In​ ​B2B​ ​too,​ ​clients​ ​operating​ ​in​ ​multiple​ ​markets​ ​or​ ​with​ ​franchise​ ​locations​ ​to manage,​ ​use​ ​sentiment​ ​technology​ ​to​ ​track​ ​opinions​ ​about​ ​suppliers,​ ​local​ ​issues​ ​and branches.​ ​This​ ​can​ ​be​ ​especially​ ​helpful​ ​in​ ​assessing​ ​service​ ​quality​ ​internally.

How​ ​does​ ​AI-powered​ ​sentiment​ ​analysis​ ​work?

Late​ ​last​ ​year,​ ​Talkwalker​ ​were​ ​the​ ​first​ ​social​ ​analysts​ ​to​ ​release​ ​proprietary​ ​image recognition​ ​technology​ ​for​ ​logos,​ ​scenes​ ​and​ ​objects,​ ​covering​ ​a​ ​vast​ ​majority​ ​of​ ​the visual​ ​web,​ ​with​ ​more than​ ​100​ ​million​ ​images​ ​processed​ ​every​ ​day​ ​across​ ​Twitter, Instagram,​ ​Facebook,​ ​online​ ​news​ ​and​ ​other​ ​sources.​ ​We​ ​​quickly​ ​realised​ ​that​ ​we wanted​ ​to​ ​apply​ ​that​ ​knowledge​ ​to​ ​sentiment​ ​technology.​ ​Using​ ​deep​ ​learning​ ​models that​ ​simulate​ ​the​ ​cognitive​ ​functions​ ​of​ ​the​ ​human​ ​brain,​ ​the​ ​technology​ ​now understands​ ​complex​ ​language​ ​patterns​ ​and​ ​entire​ ​sentences, ​and​ ​even​ ​deals​ ​with basic​ ​forms​ ​of​ ​sarcasm​ ​and​ ​irony.

While​ ​developing​ ​the​ ​algorithms​ ​that​ ​we​ ​use,​ ​we​ ​researched​ ​their​ ​success​ ​in determining​ ​accuracy.​ ​We​ ​found​ ​that​ ​the​ ​size​ ​of​ ​the​ ​training​ ​correlates​ ​with​ ​the percentage​ ​of​ ​correctly​ ​identified​ ​results.​ ​That​ ​meant​ ​the​ ​team​ ​had​ ​to​ ​classify​ ​tens of​ ​millions​ ​of​ ​results​ ​to​ ​ensure​ ​the​ ​90%​ ​accuracy​ ​rate​ ​the​ ​algorithm​ ​now​ ​delivers.

Where’s​ ​the​ ​technology​ ​headed?

Being​ ​able​ ​to​ ​accurately​ ​classify​ ​sentiment​ ​is​ ​just​ ​the​ ​start.​ ​Benchmarking​ ​brand​ ​health indicators,​ ​supplementing​ ​sentiment​ ​data​ ​with​ ​demographic​ ​information​ ​or​ ​combining​ ​it with​ ​individual​ ​product​ ​features​ ​in​ ​our​ ​platform​ ​is​ ​where​ ​the​ ​real​ ​magic​ ​happens.

Sentiment​ ​technology​ ​is​ ​fast​ ​becoming​ ​a​ ​more​ ​exact​ ​science​ ​than​ ​ever​ ​before.​ ​With​ ​the power​ ​of​ ​millions​ ​of​ ​results,​ ​accurately​ ​classified,​ ​brands​ ​are​ ​now​ ​able​ ​to​ ​take​ ​a​ ​huge step​ ​in​ ​a​ ​new​ ​direction. In​ ​a​ ​single​ ​click,​ ​brands​ ​can​ ​get​ ​the​ ​sentiment​ ​indicators​ ​they need​ ​to​ ​identify,​ ​analyse​ ​and​ ​act​ ​on​ ​insights​ ​to​ ​serve​ ​their​ ​customers​ ​well,​ ​in​ ​real​ ​time.​ ​

Have an opinion on this article? Please join in the discussion: the GMA is a community of data driven marketers and YOUR opinion counts.

Read also:

5 best ways to use social listening for your business

Digging into big data: how deep learning can unlock marketing insights