Spelen met Ngrams
NGrams wat zijn dat precies? Helaas ben ik de naam kwijt van het bedrijf - iets met ... intelligence - dat destijds een consultancy-relatie had met de Gouden Gids. De Gouden Gids had ooit een Amerikaanse eigenaar (ITT) en is sindsdien van handen verwisseld. Rond 1994 werkte Maurice De Hond er en deze wilde slimme technologie (toen al) inzetten om het zoeken te stroomlijnen. Daar kwam ik voor het eerst NGrams tegen.*
Nu dertig jaar later is het fenomeen ingeburgerd, en zie je het o.a. bij Google (books), maar dat is een iets afwijkende versie die je in taalmodellen tegenkomt. Daar gaat het om woord-"paren" die in twee (2) of drie (3) of meer constellaties achter elkaar voorkomen, dus 2-grams, of 3-grams.
Ngrams zijn belangrijk omdat losse woorden een hele andere betekenis kunnen krijgen wanneer ze in verband met iets anders worden gebruikt. New en York, of Dutch en Boy, deze laatste was mij niet bekend. Beide woorden kunnen onafhankelijk van alles betekenen, maar samen..
Deze onderstaande lijst - uit jaarverslagen - is al weer van een tijd terug, en geeft hier een voorbeeld van (N=2):
- cash flow, credit agreement, vice president, party thereto, reference amendment, income tax, benefit pension, amortization expense, property plant, segment profit, supply chain
- income expense, currency translation, executive officer**, exchange act, latin america, accounting firm, treasury stock, group consumer, health care, cross currency, discount rate, chair president
- performance incentive, audit committee, paint litigation, interest expense, purchase price, currency exchange, agent goldman, tax rate, compensation expense, check mark, mortgage company, group performance, company equity, company goldman, bank usa
- equipment net, benefit obligation, finance lease, information technology, development center, loss contingency, see note, district court, compensation plan, paper program, state court, asset retirement, reporting unit, enterprise finance, accounting officer, condition liquidity
- stock option, contribution pension, product warranty, operating cash, balance sheet, postretirement benefit, pension cost, operating lease, operating segment, specialty paint, risk management, demand environment, service cost, aerosol business, employee plan, division performance, exchange commission, interest rate, benefit cost, volume growth, employee experience, audit matter, market risk, equipment topic, trend rate, greenhouse gas, trial court, director plan, value hierarchy, pension plan, income statement, development manufacture
- business information, avenue cleveland, division paint, form k, taxonomy extension, argentine peso, product liability, customer satisfaction, bank trust, currency transaction, market value
- health safety, valuation model, growth company, leverage ratio, product line, lead pigment, lifo method,impairment interest, impairment test, impairment review, operating performance, depreciation amortization, future result, valspar acquisition, program employee, security ownership, inventory cost, value rate, climate change, information security, director independence, consumer behavior
- tax benefit, paint satisfaction, tax effect, amortization method, statement schedule, debt issuance, currency option, cost trend, benefit credit, prescription drug, management trainee, market price, debt net, consent decree,labor shortage, value topic, cybersecurity program, customer base, segment information, compensation activity, wood division, company executive, segment margin, share balance, manufacturing plant, protection agency, technology product, quality product
- service credit, division consumer, company stock, level level, life career, expertise distribution, construction period, leasing revenue, goodwill impairment, eq shareowner, growth strategy, growth rate, gibbsboro site, capacity expansion, reference form, amortization net
- region europe, peer group, gain loss, reference indenture, demand softness, swap contract, lease cost, jersey site, reporting company, equity method, restructuring expense, remediation cost, americas group, reference item, dollar amount, class action, share liability, share aggregate, settlement date, estate management
- commodity price, trademark impairment, trade practice, call provision, capital investment, caption board, caption equity, transaction price, business reputation, divestiture gain, business performance, dutch boy®, value equity, disposal group, shareholder value
- stock fund, basf se, decision maker, davie president, debt fund
- state privacy, trading policy, disability income, selling price, employee attention, employee base, end market, energy efficiency,energy use, section herein, section certification, equipment property
- equity compensation, dividend yield, cash dividend, compliance enforcement,cash source, unit level, first-out lifo, california privacy, asset impairment


Reacties